INHOPE International Association of Internet Hotlines
2007 Global Internet Trend Report
Trends associated with Illegal Content on the Internet based on the experiences of the INHOPE International Network of Internet Hotlines
September 2004 – December 2006
Version 1.0
Created by: Date: Authors:
© INHOPE 2007
INHOPE, External Services Task Group September 2007 Mr. Cormac Callanan INHOPE Dr. Nikos. P. Frydas President Safenet
This document was developed and written by the External Services Task Group of the INHOPE association – the International Network of Internet Hotlines. See www.inhope.org for further information. Thanks to Dr. Nikos P. Frydas for his work on the trend analysis Mr. Paul Durrant leader of the Statistics Working Group and all its members, Ms. Catherina Behan for her help in the preparation of this document and members of the External Services Task Group chaired by Ms. Ana-Luiza Rotta Soares and everyone for their help and support in developing this document. Please send any comments or recommendations to ceo@inhope.org marking the subject of your email as “[TREND]”. INHOPE is part funded under the European Commission Safer Internet Plan (http://ec.europa.eu/saferinternet)
Cormac Callanan, INHOPE Dublin, June 2007
© INHOPE 2007
INHOPE
2007 Global Internet Trend Report
Contents
• • • • Executive Summary Introduction INHOPE Statistics Collection Trend Results • • • • • • • Growth of the Hotline Network Scale of the problem faced by INHOPE Specific Internet Trends What happens to reports after processing Advanced Trends
Conclusion and Future Work Appendices • • • • Areas of Caution Explanation for Visible Trends INHOPE Statistics Template Statistics Primer
EXECUTIVE SUMMARY
The INHOPE International Association of Internet Hotlines is a hugely successful, respected and well known global organisation. The INHOPE hotline network has processed, and continues to process, a huge volume of reports about illegal, harmful and not-illegal content on the Internet and has, as a result, created a dynamic (time-sensitive) chart of the downside sectors of the Internet. The professional approach and the close cooperation within the network transcend international borders and ensure that criminal activity on the Internet can be successfully combated in a trans-national Internet environment. The INHOPE International Association of Internet Hotlines was established in November 1999 and stands now at 281 hotline members in 25 countries. Each of these hotlines accepts reports from the Internet public about alleged illegal content and activity on the Internet. These reports are reviewed by trained hotline personnel to determine their location on the Internet and whether the content is likely to be illegal under national law. Between September 2004 and December 2006 the INHOPE network received 900,000 reports from the public and, when added to those proactively found by the hotline personnel, the INHOPE hotline network processed a total of 1.9 million reports. These reports caused 1.25 million actions for the network. Over 162,000 reports for the whole period were forwarded to law enforcement, more than 150,000 reports were forwarded to the hosting
1
3 more hotlines/countries are in the process of joining the network
Not for publication without permission
Page 4 of 99 © INHOPE 2007
September 2007
Trend Report
Internet Service Provider, more than 170,000 reports were forwarded to the content owner and almost 33,000 reports were forwarded to another INHOPE hotline for further processing.
MONTHLY FIGURES During the last quarter of 2006 the hotline network processed an average of 91,000 reports per month having achieved an average growth rate of 2,300 reports per month. The average hotline handled approximately 3,400 reports per month. On average 35,000 of these reports were received from the public and approximately 19,000 per month were determined to refer to either illegal2 or harmful3 content. The total number of reports considered illegal or harmful by the hotlines stood at approximately 21% of the total reports processed at the last quarter of 2006. This figure is expected to keep decreasing by between 0.4 and 0.9 percent units per month. Hotlines received on average approximately 950 reports per month related to content which could not be found and 160 reports per month related to content which could not be accessed when the hotline tried to review the content. Quite interestingly an average of 6,300 reports per month (18% of the received reports) related to content which was outside the hotline remit i.e. content which the hotline was not established to accept. The INHOPE hotline network processed an average of 3,000 queries per month, while an average of 2,600 reports per month referred to Spam emails not containing illegal content. An average of 59% of reports referred to content hosted on Internet websites, 30% of reports referred to email/Spam, 1% referred to chat/messaging, 1% referred to peer-to-peer activities and 9% to other Internet services. INHOPE determined that an average of 9,600 of processed reports related to child pornography and that this number is increasing at an average of 120 reports per month. However, as a percentage of illegal or harmful
2 3
Content considered illegal in at least one country hosting an INHOPE member hotline INHOPE hotlines do not further process ‘harmful’ content once identified but accept that some users find it very disturbing
Not for publication without permission © INHOPE 2007
Page 5 of 99
content child pornography is approximately 50% and is decreasing at an average rate of 2 percent units per year. INHOPE determined that an average of 220 reports per month related to racism or xenophobia and this number is increasing by 6 reports per month. This number represents 1% of the illegal or harmful content and is increasing at 24% per year. The hotline network needed to take an average of 46,000 actions per month. This was increasing at 800 actions per month. An average of 6,400 reports per month were forwarded to law enforcement for investigation The INHOPE International Association of Internet Hotlines is a very successful initiative and brings substantial levels of support and comfort to Internet users harmed by criminal activity on the Internet. The network is substantially funded through the EU Safer Internet Action Plan and greater resources are urgently required if this fight is to continue especially in the face of growing visibility and growing numbers of reports being processed. Finally, we should be careful when attempting to explain the observed patterns of the time series. While causation implies correlation, correlation does not imply causation.
Page 6 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
INTRODUCTION
PURPOSE This report presents the statistics collected by INHOPE since September 2004 through December 2006 from the individual INHOPE member hotlines and identifies clear trends which these collective statistics portray. This report does not offer any interpretation about the cause of these statistics and trends. This is the first trend report from the INHOPE International Association of Internet Hotlines4. The specific trends to be quoted were determined during the analysis phase and proven as statistically significant. A selected subsection of the wide range of data which INHOPE collects was targeted during this first phase of analysis.
TARGET AUDIENCES The primary audience of this report are INHOPE members so that the international trends identified can be verified against local experience and results. The second key audience includes stakeholders who have been asked to consider these issues at a national and international level. These stakeholders include EC personnel and politicians, Internet industry
including fixed and wireless/mobile, national governments, trans-national governmental organisations, child welfare and child rights organisations,
4
in some countries Internet hotlines are referred to as Internet Tiplines
Not for publication without permission © INHOPE 2007
Page 7 of 99
national and international law enforcement, media, Internet users in each country and the general public. The report will be published on the INHOPE website and distributed to INHOPE’s regular contacts including national and international print, online and broadcast media as appropriate.
EXCLUDED FROM REPORT The report does not provide detailed individual statistics or trend analysis for any single hotline or country since this is left to those hotlines to produce and distribute. This first trend report does not yet offer reasons why the trends identified exist. There are many reasons which might be the cause of such trends and further research will be conducted to shed light on these issues.
Page 8 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
INHOPE STATISTICS COLLECTION
INHOPE hotlines collect a range of information in relation to the day-to-day work of the hotline. The minimum data collected is agreed by the INHOPE General Assembly and is specified in a Best Practice Paper on Statistics. The General Assembly adopted this Best Practice Paper and specified that these statistics be submitted to the INHOPE online member’s statistics database on a monthly basis. For example, such data includes the number of reports received about child pornography, the number of queries received, the number of such reports which were determined by the hotline to be child pornography, the total number of reports received, the total reports forwarded to law enforcement, etc. INHOPE does not exist to collect this type of trend information. INHOPE exists to coordinate a network of hotlines which respond to illegal content and activity on the Internet. Of course, in the process of performing this work INHOPE collects data about the work activities of the hotlines. This data has now been analysed to produce the trends included in this report. The purpose of the INHOPE template is to use a consistent method of measuring the activities of a hotline. Hotlines can gather any additional statistics they choose but the INHOPE template determines the minimum required by INHOPE. A key part of this process is to ensure as far as possible that different organisations from different languages and cultures submit data using the
Page 9 of 99
Not for publication without permission © INHOPE 2007
same understandings and definitions and collect the statistics under essentially the same conditions. The INHOPE template is extremely useful in measuring the workload of a hotline and this information is important to INHOPE and the individual INHOPE members. It is important to note that the design does not directly measure the scale of the problem of illegal and harmful use of the Internet. INHOPE is aware of increasing interest and demand for information about the scale and trends of illegal and harmful use of the Internet as reflected in the knowledge and experience gained by INHOPE and INHOPE members. Further details and definitions are outlined in Appendix #III
Page 10 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
TREND RESULTS
The statistics below have been analysed. The chosen statistics form a subset of the total available statistics. They are designed to represent what INHOPE believes external stakeholders would be interested in learning. 1. 2. Growth of the INHOPE International Association of Internet Hotlines and contributing to INHOPE statistics database Scale of the Internet problem faced by INHOPE 2.1. Total Reports Processed Total Reports Processed per Hotline 2.2. Total Number of Reports Received Total Number of Reports Received per Hotline 2.3. Total Reports about Illegal or Harmful Content 2.4. Total Unique Actionable Reports 2.5. Reports considered Not Illegal or which could Not be Determined 2.5.1. 2.5.2. 2.5.3. 2.5.4. Spam Not Containing Illegal Content Other Content5 which is not Illegal Queries Processed Outside the Hotline Remit Not Found Not Accessible
2.6. Classification of Reports by Internet Service 3. The INHOPE Network of Hotlines specific Internet trends: 3.1. Child Pornography and % 3.2. Other Child-Related Content and % 3.3. Racism & Xenophobia and % 3.4. Extreme Adult Content and %
5
This refers to content other than content “considered to be potentially illegal or harmful” or “SPAM not containing illegal content”.
Page 11 of 99 Not for publication without permission © INHOPE 2007
4.
What happens to reports once they have been determined to be illegal? 4.1. Total Resulting Hotline Actions 4.1.1. 4.1.2. 4.1.3. 4.1.4. Reports Transmitted to Law Enforcement Reports Transmitted to Hosting ISP Reports Forwarded to another INHOPE Hotline Other Resulting Actions
Page 12 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
NOTES
1. 2. 3.
The Unit of Time is, one month, i.e. the statistics collected are monthly. The statistics in this report are sums over all contributing INHOPE hotlines. The percentage (%) statistics are obtained from the corresponding absolute numbers divided by the appropriate total number of reports, at the corresponding unit of time. Some statistics are sums of other statistics. When appropriate, results quoted have been tested for statistical significance. The actual measurements are represented by the points joined with the dotted line (all in red). The smooth curve (in blue) represents the estimated trend line. The straight line (in green) represents the linear trend whose slope gives the average monthly rate of change, quoted above. Significant seasonal variations were manually removed for the trend calculations and replaced by interpolated data which were smoothed using “exponential smoothing”. Six distinct criteria were used to test the statistical significance of the results and to select the best of four trend models. Trend lines for statistics which are functions (sums, or ratios) of other statistics are formed as the same functions of the corresponding trend line polynomials. All statistical results are rounded to two significant digits, with the exception of the maximum values which are rounded to three significant digits. Monthly rates of change, quoted as %, are the ratio of the rate of change per month (in reports per month, etc) over the value of the statistic at the last quarter of 2006 (three-month trend average). For example, if the rate of change is 2,500 more reports per month, while the value of the statistic at the last quarter of 2006
4. 5. 6. 7. 8. 9.
10.
11.
12.
Page 13 of 99
Not for publication without permission © INHOPE 2007
(three-month trend average) is 50,000, the % monthly rate of change is 2,500/50,000 = +5.0% per month, or 12x5% = +60% per year. This form of the rate of change is more informative about the speed at which a certain statistic changes with time. As another example, if the rate of change is 2% fewer reports per month, while the value of the statistic at the last quarter of 2006 (three-month trend average) is 50%, the % monthly rate of change is -2%/50% = -4.0% per month, or 12x(-4)% = -48% per year. 13. Any rate of change which is smaller than 0.1% per month, is considered “practically nil”.
Page 14 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
1.
NUMBER OF CONTRIBUTING HOTLINES
OVERVIEW This statistic shows the number of independent organisations contributing to the INHOPE statistics that follow. INHOPE has taken measures to ensure that individual hotlines collect and classify reports under essentially the same conditions. This is necessary for many reasons and, especially within the context of the current report, for collating hotline statistics. The hotlines contributing to the INHOPE statistics are of course members of INHOPE, but their number is less than the size of INHOPE membership. This is so because between admission of an organisation to INHOPE and production of report statistics by that organisation a number of months can elapse. TREND The number of contributing hotlines which are members of INHOPE exhibits an increasing trend which, between September 2004 and December 2006, stood at an average of one new hotline every three months. The number of contributing hotlines at the end of the observation period (December 2006) stood at 27, while the membership total now stands at 28 hotlines (September 2007).
Page 15 of 99
Not for publication without permission © INHOPE 2007
30 25 20 15 10 5 0 04-09
04-12
05-03
05-06
05-09
05-12
06-03
06-06
06-09
06-12
07-03
Number of Contributing INHOPE Hotlines
The graph above shows the time at which new hotlines, members of INHOPE, start the production of reports and associated statistics. INHOPE COMMENT The INHOPE network dedicates significant efforts to identifying and encouraging new hotlines to develop and grow. Dedicated INHOPE membership resources operate to ensure the new initiatives are supported and guided towards best international practice in the operation of Internet hotlines. The membership process is complex, comprehensive and exhaustive. Significant efforts are spent by INHOPE and the prospective member to verify the level of stakeholder support for the new hotline. Current resources support the current level of inclusion of new members. INHOPE would need additional resources if the rate of membership inclusion increases. The current rate at which new membership occurs (one new productive hotline per quarter) is low if one considers the UN membership, but impressive if one considers the very elaborate procedure by which a candidate hotline is admitted to INHOPE. To this, one should acknowledge the effort required by the new hotline to launch the mechanism by which reports are generated, collected, processed and classified.
Page 16 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
2. SCALE OF THE INTERNET PROBLEM FACED BY INHOPE
OVERVIEW Reports received by hotlines provide information about the Internet service and content which is being reported and additional descriptive data which indicates what the Internet user believes to be the nature of illegality associated with the reported content (“allegation”). INHOPE hotlines obtain their reports via three sources: 1. From the public (external reports) 2. Derived from the external reports (derived reports) 3. Via the hotline proactive action (proactive reports) The total number of reports processed by an INHOPE hotline is exactly the sum of the above and they are called hotline-determined reports. Hotline-determined statistics are very important because they tell us what actually happens, as opposed to what the reporting public believes to be the situation. Although not frequently, members of the public sometimes mistake one category for another. For example, child pornography is overestimated by ~18%6, while adult pornography is underestimated by ~8%7. Once such reports are processed by hotline personnel, a determination is decided based on hotline knowledge, training and expertise in relation to
6 7
See Section 5 Advanced Trends See Section 5 Advanced Trends
Not for publication without permission © INHOPE 2007
Page 17 of 99
the reported content (“hotline determination”). This determination is a more accurate reflection of the nature of the illegal content. By recording the “allegation” and the “hotline determination” INHOPE can see the view as perceived by the reporting public in comparison to the hotline determination. This section of the trend report compares the hotline determination of illegal content with that reported by members of the public.
Page 18 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
2.1 TOTAL REPORTS PROCESSED
OVERVIEW “Total Reports Processed” is the sum of all reports the hotlines process. This sum comprises the external reports, which originate from outside the hotline (mainly from the general public), the derived reports, which are generated by follow-up actions upon external reports and the proactive reports, which originate from hotline proactive actions. This is one of the most important statistics, because it is a measure of the total volume of work carried out by the hotlines.
300,000
300,000 300,000
250,000 200,000 150,000 100,000
250,000 250,000 200,000 200,000 150,000 150,000 100,000 100,000 50,000 50,000 00
0501 0501
0504 0504
0507 0507
0510 0510
0601 0601
0604 0604
0607 0607
0610 0610
Illegal Illegal
Spam Spam
Other Other
Queries Queries
Not accessible Not accessible
Not found Not found
Outside Outside
50,000 0
0501
0504
0507
0510
0601
0604
0607
0610
Illegal
Spam
Other
Queries
Not accessible
Not found
Outside
Total number of Reports Processed (by the INHOPE network) classified by category Trends and Actual (inset)
TREND The number of “Total Reports Processed” stood at approximately 91,000 reports per month, at the last quarter of 2006. The maximum processed in
Page 19 of 99 Not for publication without permission © INHOPE 2007
any one month was 262,000 reports. The average rate of increase, between September 2004 and December 2006, was 2,300 additional reports per month (equivalently, +31% per year). The expectation for the beginning of 2007 is 2,000 - 2,700 additional reports per month (95% confidence interval).
300,000 250,000 200,000 150,000 100,000 50,000 0 04-09
04-12
05-03
05-06
05-09
05-12 06-03
06-06
06-09
06-12
07-03
Total number of Reports Processed
TREND The number of “Total Reports Processed per Hotline” stood at approximately 3,400 reports per month, at the last quarter of 2006. The maximum processed in any one month was 10,500 reports. The average rate of increase, between September 2004 and December 2006, was 67 additional reports per month (equivalently, +24% per year). The expectation for the beginning of 2007 is 55 - 80 additional reports per month (95% confidence interval).
12,000 10,000 8,000 6,000 4,000 2,000 0 04-09
04-12 05-03
05-06 05-09
05-12
06-03 06-06
06-09 06-12
07-03
Total number of Reports Processed per Hotline
•
The actual measurements are represented by the points joined with the dotted line (all in red).
Not for publication without permission
Page 20 of 99 © INHOPE 2007
September 2007
Trend Report
• • •
The smooth curve (in blue) represents the estimated trend line. The straight line (in green) represents the linear trend whose slope gives the average monthly rate of change, quoted above. See also “NOTES” in p. 13.
INHOPE COMMENT The number of Total Reports Processed is, perhaps, the best measure of the volume of the work carried out by the INHOPE hotlines. Removing any seasonal data, INHOPE hotlines process ~700 reports per working hour8, while this workload increases at an average rate of 30% per year! Of course the quoted increase in the workload of the INHOPE network is also due to the concurrent increase in the number of hotlines. To count for that effect, one may use the average number of reports processed per hotline. The average hotline processes ~26 reports per working hour, while this workload increases at an average rate of 24% per year! The volume of work for each hotline is increasing significantly at 2%-3% per month. This increase dictates the number of hotline staff and resources required to be in place to respond to these reports. Hotline content analysts incur the significant workload caused by these additional reports. If one assumes that each report requires between 5 and 20 minutes of work, one concludes that the average hotline requires between 3 and 10 analysts. The above mentioned annual growth rates are impressive and one suspects that available resources do not grow at the same pace, which indicates improved hotline efficiency during the time period covered by this report. The increase in reports could be due to many reasons which cannot be determined by the analysis performed for this report. These reasons could include: • • • •
8
Change in hotline numbers Change in hotline visibility Change in hotline activity (derived, proactive) Change in criminal activity
A total workload of 3,400 reports per month, assuming 19 working days per month (5 days/week, and 10.5 work-months/year for a full-time salaried person) and 7 work-hours per day is equivalent to 25.9 reports/working-hour.
Not for publication without permission © INHOPE 2007
Page 21 of 99
•
Change in visibility of criminal activity
It is suspected that the root cause for the observed growth in the Number of Reports Processed per Month is a combination of the above and possibly other reasons. More research needs to be done on this very challenging topic.
Page 22 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
2.2 TOTAL REPORTS RECEIVED
OVERVIEW This statistic shows the total number of reports received per month by all contributing INHOPE hotlines. These reports have been produced by the general public and they generate crucial statistics because they are a measure of the visibility of the INHOPE network and, in particular, of the determination of Internet users to act against illegal content in general.
80,000 80,000
80,000 70,000 60,000 50,000 40,000 30,000 20,000 10,000 0
70,000 70,000 60,000 60,000 50,000 50,000 40,000 40,000 30,000 30,000 20,000 20,000 10,000 10,000 00
0501 0501
0504 0504
0507 0507
0510 0510
0601 0601
0604 0604
0607 0607
0610 0610
Illegal Illegal
Spam Spam
Other Other
Queries Queries
Outside Outside
0501
0504
0507
0510
0601
0604
0607
0610
Illegal
Spam
Other
Queries
Outside
Total number of Reports Received, classified by category – Trends and Actual (inset)
TREND The number of “Total Reports Received” stood at approximately 35,000 reports per month, at the last quarter of 2006. The maximum processed in any one month was 75,000 reports. The average rate of increase, between September 2004 and December 2006, was 520 additional reports per
Page 23 of 99
Not for publication without permission © INHOPE 2007
month (equivalently, +18% per year). The expectation for the beginning of 2007 is 450 - 590 additional reports per month (95% confidence interval).
80,000 70,000 60,000 50,000 40,000 30,000 20,000 10,000 0 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03
Total number of Reports Received
TREND The number of “Total Reports Received per Hotline” stood at approximately 1,200 reports per month, at the last quarter of 2006. The maximum processed in any one month was 3,000 reports. There was no statistically significant monthly change, between September 2004 and December 2006 and none is expected for the beginning of 2007.
3,000 2,500 2,000 1,500 1,000 500 0 04-09 04-12
05-03 05-06 05-09
05-12 06-03 06-06 06-09
06-12 07-03
Total Number of Reports Received per Hotline
• • • •
The actual measurements are represented by the points joined with the dotted line (all in red). The smooth curve (in blue) represents the estimated trend line which, in this case, is cubic. The straight line (in green) represents the linear trend whose slope gives the average monthly rate of change, quoted above. See also “NOTES” in p. 13.
Page 24 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
INHOPE COMMENT The volume of reports received by each hotline is increasing by over 1% per month. This increase dictates the number of hotline staff and resources required to be in place to respond to these reports. Hotline content analysts incur the significant workload caused by these additional reports. The Total number of Reports Received is, perhaps, the best measure of the visibility of an INHOPE hotline. Removing any seasonal data, INHOPE hotlines receive ~270 reports per working hour, while this workload increases at an average rate of 18% per year! On the other hand, the average hotline receives ~9 reports per working hour, a figure which does not alter over time. The increase in reports could be due to many reasons which cannot easily be determined by the analysis performed for this report. These reasons could include: • • • • • Change in hotline numbers Change in hotline visibility Change in hotline activity (derived, proactive) Change in criminal activity Change in visibility of criminal activity
To determine the underlying cause of this increase in the number of reports processed by the INHOPE network more research needs to be done.
Page 25 of 99
Not for publication without permission © INHOPE 2007
2.3
TOTAL REPORTS ABOUT ILLEGAL OR HARMFUL CONTENT9
OVERVIEW “Total Reports about Illegal or Harmful Content” groups together all content categories that are potentially illegal in at least one INHOPE country. The current section compares the total number of reports which have been considered to be illegal by members of the public (the “Received Reports”) to those which the INHOPE network of hotlines determines to be illegal (the “Processed Reports”).
30,000 30,000 25,000 25,000 20,000 20,000
30,000 25,000 20,000 15,000 10,000 5,000 0
15,000 15,000 10,000 10,000 5,000 5,000 00
0501 0501
0504 0504
0507 0507
0510 0510
0601 0601
0604 0604
0607 0607
0610 0610
CP CP
Other child cont. Other child cont.
Porn Porn
Extreme Extreme
Other illegal Other illegal
0501
0504
0507
0510
0601
0604
0607
0610
CP
Other child cont.
Porn
Extreme
Other illegal
Total number of Processed Reports about Illegal or Harmful Content classified by category – Trends and Actual (inset)
9 Potentially illegal or harmful content comprises reports about Child Pornography, Child Trafficking, Child Sex Tourism, Child Nudism, Child Grooming Activities,, Child Erotica/Inappropriate Images of Children, Adult Pornography accessible to Children, Extreme Adult Pornography, Adult Pornography, Racism and Xenophobia, Promoting Violence Against an Individual, Terrorism and Drugs (not all these are illegal in every country).
Page 26 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
TREND The number of “Total Received Reports about Illegal or Harmful Content”10 stood at approximately 19,000 reports per month, at the last quarter of 2006. The maximum processed in any one month was 32,200 reports. The average rate of increase, between September 2004 and December 2006, was 200 additional reports per month (equivalently, +12% per year). The expectation for the beginning of 2007 is 130 - 270 additional reports per month (95% confidence interval). As a percentage of all received (external) reports, the number of “Total Received Reports about Illegal or Harmful Content” stood at approximately 61% at the last quarter of 2006. The maximum achieved in any one month was 75.3%. Between September 2004 and December 2006 it decreased slowly at an average rate of 0.18 percent units per month (equivalently, -3.6% per year). The expectation for the beginning of 2007 is 0.12 – 0.24 fewer percent units per month (95% confidence interval).
35.000 30.000 25.000 20.000 15.000 10.000 5.000 0 04-09
80% 80% 70% 70% 60% 60% 50% 50% 40% 40% 30% 30% 20% 20% 10% 10% 0% 0%
04-12 05-03
05-06 05-09
05-12
04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03 04-09 04-12 05-12 06-03 06-03 06-06 05-03 05-06 05-0906-12 06-06 06-09 06-12 07-03 06-09 07-03
Total number of Received Reports about Illegal or Harmful Content and as a percentage of all reports received
TREND The number of “Total Processed Reports about Illegal or Harmful Content”11 stood at approximately 19,000 reports per month, at the last quarter of 2006. The maximum processed in any one month was 27,500 reports. The average rate of increase, between September 2004 and December 2006, was 280 additional reports per month (equivalently, +18% per year). The
10 11
It is reminded that these are reports received from the general public.
It is reminded that these reports are determined by the hotlines
Not for publication without permission © INHOPE 2007
Page 27 of 99
expectation for the beginning of 2007 is 250 - 300 additional reports per month (95% confidence interval). As a percentage of all processed (hotline-determined) reports, the number of “Total Processed Reports about Illegal or Harmful Content” stood at approximately 21% at the last quarter of 2006. The maximum achieved in any one month was 45.6%. Between September 2004 and December 2006 it decreased at an average rate of 0.62 percent units per month (equivalently, -36% per year). The expectation for the beginning of 2007 is 0.38 – 0.87 fewer percent units per month (95% confidence interval).
50% 50% 45% 45% 40% 40% 35% 35% 30% 30% 25% 25%
30,000 25,000 20,000 15,000 10,000 5,000
20% 20% 15% 15% 10% 10% 5% 5% 0% 0% 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03
0 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03
Total number of Processed Reports about Illegal or Harmful Content and as a percentage of all reports processed (inset)
• • • •
The actual measurements are represented by the points joined with the dotted line (all in red). The smooth curve (in blue) represents the estimated trend line. The straight line (in green) represents the linear trend whose slope gives the average monthly rate of change, quoted above. See also “NOTES” in p. 13. Potentially illegal or harmful content comprises reports about Child Pornography, Child Trafficking, Child Sex Tourism, Child Nudism, Child Grooming Activities,, Child Erotica/Inappropriate Images of Children, Adult Pornography accessible to Children, Extreme Adult Pornography, Adult Pornography, Racism and Xenophobia, Promoting Violence Against an Individual, Terrorism and Drugs (not all these are illegal in every country).
•
Page 28 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
INHOPE COMMENT Reports considered illegal or harmful are the main reason for the generation of reports in the first place: Such reports may be crucial in stopping criminal acts, or may be used in charting illegal or harmful content on the Internet. The INHOPE network receives 19,000 reports per month with content believed by the general public to fall in the category of potentially illegal or harmful, as determined above, and this number increases at ~13% per year. By coincidence, the INHOPE network also determines 19,000 reports per month with content falling in the category of potentially illegal or harmful, as determined above, but this number increases faster than for external reports, at ~18% per year. The fact that both the external and the hotline-determined ‘illegal’ reports per month stood at ~19,000 at the 4th quarter of 2006 is a coincidence: In addition to reports received by the public, hotlines generate reports proactively, or derive new reports from the external ones. The increase in reports could be due to many reasons which cannot readily be determined by the analysis performed for this report. These reasons could include: • • • • • • Change in hotline reports Change in knowledge and expertise of those making reports creating more accurate reports about illegal/harmful content Change in hotline visibility Change in hotline activity (derived, proactive) Change in criminal activity Change in visibility of criminal activity
To determine the underlying cause of this increase in the number of reports processed by the INHOPE network more research needs to be done.
Page 29 of 99
Not for publication without permission © INHOPE 2007
2.4
TOTAL UNIQUE ACTIONABLE REPORTS12
OVERVIEW “Actionable Reports” are the reports the hotlines take action on, because they are likely to be illegal under national Law. As a result, the hotlines remove any reports outlined in section 2.5. From the set of “Actionable Reports” defined above, each hotline seeks to eliminate any duplicate reports to end with the subset of “Total Unique Actionable Reports”. It is important to note that there is currently no facility for INHOPE members to detect and eliminate duplicate reports across the INHOPE network. “Total Unique Actionable Reports” is a measure of the volume of what each country defines as illegal content and this is determined by the INHOPE hotline network. Each of these reports requires action (or sometimes multiple actions) by the local INHOPE hotline. TREND The number of “Total Unique Actionable Reports” stood at approximately 19,000 reports per month, at the last quarter of 2006. The maximum processed in any one month was 26,300 reports. The average rate of increase, between September 2004 and December 2006, was 210 additional reports per month (equivalently, +13% per year). The expectation for the beginning of 2007 is 76 - 340 additional reports per month (95% confidence interval).
12
These are reports unique per hotline, but not necessarily unique within INHOPE.
Not for publication without permission
Page 30 of 99 © INHOPE 2007
September 2007
Trend Report
As a percentage of all processed (hotline-determined) reports, the number of “Total Unique Actionable Reports” stood at approximately 21% at the last quarter of 2006. The maximum achieved in any one month was 54.0%. Between September 2004 and December 2006 it decreased at an average rate of 0.96 percent units per month (equivalently, -54% per year). The expectation for the beginning of 2007 is 0.62 – 1.30 fewer percent units per month (95% confidence interval).
27.000 24.000 21.000 18.000 15.000 12.000 9.000 6.000 3.000 0 04-09
60% 60% 50% 50% 40% 40% 30% 30% 20% 20% 10% 10% 0% 0% 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03
04-12 05-03
05-06 05-09
05-12
06-03 06-06
06-09 06-12
07-03
Number of Total Unique Actionable Reports. and as a percentage of all reports processed (inset)
• • • •
The actual measurements are represented by the points joined with the dotted line (all in red). The smooth curve (in blue) represents the estimated trend line. The straight line (in green) represents the linear trend whose slope gives the average monthly rate of change, quoted above. See also “NOTES” in p. 13.
INHOPE COMMENT Assuming 15 minutes per such report, at least 36 staff working without break is required to deal with this volume. If one takes into account the fact that the flow of such reports is not uniform over time (i.e. there are fluctuations) while these reports need to be processed immediately if they are to be useful at all, it is evident then that the number of staff must be able to cope with the peaks and not with the average. From the graph above one concludes that a safe estimate of the max workload per month is around 40% above the 19,000 mark, requiring about 51 analysts, or just less than two analysts for the average hotline. Even this
Page 31 of 99 Not for publication without permission © INHOPE 2007
figure is an underestimate because the reports must be processed within the day, or at least the next day (under exceptionable circumstances), and not within the month. Hence, the fluctuations of the number of Unique Actionable Reports per day are expected to have more dramatic peaks. In addition to the above, the total number of analysts required in practice is even higher because small hotlines are more likely to have to use more analysts than their share of this workload requires: For example, if the workload dictates 1.5 analysts the hotline will have to employ two (complete!) persons.
Page 32 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
2.5 REPORTS NOT CONSIDERED TO BE ILLEGAL OR HARMFUL
OVERVIEW This section describes the specific trends relating to content not considered illegal, as determined by INHOPE hotlines and classified into several distinct categories. These categories are: 1. Total Reports about SPAM Not Containing Illegal Content 2. Total Reports about Other Content13 3. Total number of Queries Processed 4. Total Reports Outside the Hotline Remit, content which has Not been Found and content which is Not Accessible
65%
3%
21% 7%
3.1% 1.0% 0.2%
Illegal
Spam
Other
Queries
Not accessible
Not found
Outside
Categories, and their contribution to the Total number of Reports Processed (average of trend at the last quarter of 2006)
The purpose of this section is to highlight the volume of work which hotlines incur relating to non-illegal content. Each of these reports still requires
13
Other than content “concidered to be potentially illegal or harmful”, or “Spam not containing illegal content”
Not for publication without permission © INHOPE 2007
Page 33 of 99
some hotline attention such as recording the reports in the hotline database, determining that the content is clearly not illegal and/or outside the hotline remit, responding to the reporting person indicating this decision and recording these actions taken in the hotline database. These reports then need to be classified and submitted for the INHOPE data collection process.
Page 34 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
2.5.1 REPORTS ABOUT SPAM NOT CONTAINING ILLEGAL CONTENT
OVERVIEW “Reports about SPAM Not Containing Illegal Content” are the reports processed by the hotlines about unsolicited emails - which do NOT contain illegal content. Hotlines do respond to these reports and the number of such reports contributes to the volume of work performed by the INHOPE hotline network. TREND The number of “Processed Reports about SPAM Not Containing Illegal Content” stood at approximately 2,600 reports per month, at the last quarter of 2006. The maximum processed in any one month was 11,100 reports. The average rate of decrease, between September 2004 and December 2006, was 120 fewer reports per month (equivalently, -57% per year). The expectation for the beginning of 2007 is 71 - 170 fewer reports per month (95% confidence interval). As a percentage of all processed reports, the number of “Processed Reports about SPAM Not Containing Illegal Content” stood at approximately 2.9% at the last quarter of 2006. The maximum achieved in any one month was 22.5%. Between September 2004 and December 2006 it decreased fast at an average rate of 0.57 percent units per month (equivalently, -230% per year). The expectation for the beginning of 2007 is 0.53 – 0.60 fewer percent units per month (95% confidence interval), but obviously this rate of fall will soon ease-off.
Page 35 of 99
Not for publication without permission © INHOPE 2007
30% 30%
14.000 12.000 10.000 8.000 6.000 4.000 2.000
25% 25% 20% 20% 15% 15% 10% 10% 5% 5% 0% 0% 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03
0 04-09
04-12 05-03
05-06 05-09
05-12
06-03 06-06
06-09 06-12
07-03
Number of Processed Reports about SPAM Not containing Illegal Content and as a percentage of the total reports processed (inset)
• • • •
The actual measurements are represented by the points joined with the dotted line (all in red). The smooth curve (in blue) represents the estimated trend line. The straight line (in green) represents the linear trend whose slope gives the average monthly rate of change, quoted above. See also “NOTES” in p. 13.
INHOPE COMMENT This is one of a small number of INHOPE statistics of absolute numbers (as opposed to percentages) whose numbers are decreasing with time, in spite the influx of more reports per month. Examining the graph one sees that there are two distinct periods: Up to Feb 2006 there were 4-6,000 reports/month, while thereafter there was a sharp drop to 1,500-2,500 reports/month.
Page 36 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
2.5.2 REPORTS ABOUT OTHER CONTENT OR CONTENT NOT CONSIDERED
ILLEGAL
OVERVIEW “Reports about Other Content or Content Not Considered Illegal” are the reports processed by the hotlines which are not considered to be “illegal or harmful” nor “spam not containing illegal content”. These reports are often generated by Internet users who are frustrated with specific content on the Internet and do not know where else, apart from hotlines, such content can be reported. They themselves often already accept that the content is unlikely to be illegal but still want to report it to an ‘authority’ for review. TREND The number of “Processed Reports about Other Content or Content which is Not Illegal” stood at approximately 62,000 reports per month, at the last quarter of 2006. The maximum processed in any one month was 231,000 reports. The average rate of increase, between September 2004 and December 2006, was 1,800 additional reports per month (equivalently, +35% per year). The expectation for the beginning of 2007 is 1,100 – 2,500 additional reports per month (95% confidence interval). As a percentage of all processed reports, the number of “Processed Reports about Other Content or Content which is Not Illegal” stood at approximately 65% at the last quarter of 2006. The maximum achieved in any one month was 88.0%. Between September 2004 and December 2006 it increased at an average rate of 1.1 percent units per month (equivalently, +21% per year). The expectation for the beginning of 2007 is 0.79 – 1.50 additional
Page 37 of 99
Not for publication without permission © INHOPE 2007
percent units per month (95% confidence interval), but obviously the growth rate will have to ease-off soon.
250,000 200,000 150,000 100,000 50,000 0 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03
Number of Processed Reports about Other Content or Content which is Not Illegal and as a percentage of the total reports processed.. • • • • The actual measurements are represented by the points joined with the dotted line (all in red). The smooth curve (in blue) represents the estimated trend line. The straight line (in green) represents the linear trend whose slope gives the average monthly rate of change, quoted above. See also “NOTES” in p. 13.
100% 100% 90% 90% 80% 80% 70% 70% 60% 60% 50% 50% 40% 40% 30% 30% 20% 20% 10% 10% 0% 0% 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03
INHOPE COMMENT This is a category with a remarkable combination of strong growth rate (+21%/year), for the % statistic, which is already at a very high point: 65% of all processed reports! It is evident that in the next trend report this category needs to be analysed to smaller sub-categories.
Page 38 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
2.5.3 QUERIES PROCESSED
OVERVIEW “Queries Processed” are ‘reports’ which are not in fact reports but are queries relating to the illegal or harmful use of the Internet. Hotlines respond to these queries and their number contributes to the volume of work performed by the INHOPE network. TREND The number of “Queries Processed” stood at approximately 3,000 queries per month, at the last quarter of 2006. The maximum processed in any one month was 4,170 queries. The average rate of increase, between September 2004 and December 2006, was 74 additional queries per month (equivalently, +30% per year). The expectation for the beginning of 2007 is 54 – 94 additional queries per month (95% confidence interval). As a percentage of all processed reports, the number of “Queries Processed” stood at approximately 3.1% at the last quarter of 2006. The maximum achieved in any one month was 5.74%. There was no statistically significant monthly change, between September 2004 and December 2006 and none is expected for the beginning of 2007.
Page 39 of 99
Not for publication without permission © INHOPE 2007
5,000 4,500 4,000 3,500 3,000 2,500 2,000 1,500 1,000 500
6% 6% 5% 5% 4% 4% 3% 3% 2% 2% 1% 1% 0% 0% 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03
0 04-09
04-12
05-03
05-06
05-09
05-12
06-03
06-06
06-09
06-12
07-03
Number of Queries Processed and as a % of the total reports processed.
• • • •
The actual measurements are represented by the points joined with the dotted line (all in red). The smooth curve (in blue) represents the estimated trend line. The straight line (in green) represents the linear trend whose slope gives the average monthly rate of change, quoted above. See also “NOTES” in p. 13.
INHOPE COMMENT INHOPE hotlines receive a wide range of queries about content and activity on the Internet. Sometimes this relates to parents asking for help with their own children’s activities on the Internet and sometimes it relates to the legality/illegality of content they have seen on the Internet. These reports indicate the desire by such persons to have access to a professionally-styled helpline with knowledge of Internet content issues and the legislation in each country. INHOPE provides a valuable service to such persons by listening and recording their concerns and then, when possible, offering experienced advice or knowledge on their area of query. This activity also consumes substantial hotline resources. Usually more resources are required for this type of report because of the need for direct communication with the reporting person and the need to be sensitive to the reporter’s motivations for reporting the content in the first place.
Page 40 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
2.5.4 REPORTS OUTSIDE THE HOTLINE REMIT, CONTENT WHICH HAS NOT
BEEN FOUND AND CONTENT WHICH IS NOT ACCESSIBLE
OVERVIEW This category comprises the following types of reports: 1. Reports which are outside the hotline remit (because the hotline does not deal with the reported material). 2. Reports about content which the hotline could not find on the Internet from the report received (because the report was not accurate, or because the content is not reachable, or because the content has been moved by the content owner or already removed by the content hosting organisation etc). The content might have been illegal when reported (or not) but since an assessment cannot be made by the hotline these reports cannot be processed any further. 3. Reports which are not accessible at the reported Internet address (because they are in a closed member’s area, or password protected, or encrypted, etc). TREND The number of “Reports Outside the Hotline Remit” stood at approximately 6,300 reports per month, at the last quarter of 2006. The maximum processed in any one month was 34,500 reports. The average rate of increase, between September 2004 and December 2006, was 240 additional reports per month (equivalently, +47% per year). The
Page 41 of 99
Not for publication without permission © INHOPE 2007
expectation for the beginning of 2007 is 150 – 340 additional reports per month (95% confidence interval). As a percentage of all processed reports, the number of “Reports Outside the Hotline Remit” stood at approximately 6.8% at the last quarter of 2006. The maximum achieved in any one month was 22.3%. Between September 2004 and December 2006 it increased at an average rate of 0.24 percent units per month (equivalently, +43% per year). The expectation for the beginning of 2007 is 0.17 – 0.32 additional percent units per month (95% confidence interval).
25% 25%
40.000 35.000 30.000 25.000 20.000 15.000 10.000 5.000
20% 20% 15% 15% 10% 10% 5% 5% 0% 0% 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03
0 04-09
04-12 05-03
05-06 05-09
05-12
06-03 06-06
06-09 06-12
07-03
Number of reports Outside the Hotline Remit and as a percentage of the total reports processed (inset)
TREND The number of reports referring to content “Not Found” stood at
approximately 950 reports per month, at the last quarter of 2006. The maximum processed in any one month was 10,800 reports. The average rate of decrease, between September 2004 and December 2006, was 9.0 fewer reports per month (equivalently, -11% per year). The expectation for the beginning of 2007 is 0.7 – 17.0 fewer reports per month (95% confidence interval). As a percentage of all processed reports, the number of reports referring to content “Not Found” stood at approximately 1.0% at the last quarter of 2006. The maximum achieved in any one month was 19.2%. Between September 2004 and December 2006 it decreased fast at an average rate
Page 42 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
of 0.096 percent units per month (equivalently, -9.2% per month). The expectation for the beginning of 2007 is 0.091 – 0.100 fewer percent units per month (95% confidence interval), but obviously this rate will ease-off soon.
12.000 10.000 8.000 6.000 4.000 2.000 0 04-09
20% 20% 18% 18% 16% 16% 14% 14% 12% 12% 10% 10% 8% 8% 6% 6% 4% 4% 2% 2% 0% 0% 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03
04-12 05-03
05-06 05-09
05-12
06-03 06-06
06-09 06-12
07-03
Number of reports referring to content Not Found and as a percentage of the total reports processed (inset)
TREND The number of reports referring to content which is “Not Accessible” stood at approximately 160 reports per month, at the last quarter of 2006. The maximum processed in any one month was 7,480 reports. The average rate of decrease, between September 2004 and December 2006, was 38 fewer reports per month (equivalently, -23% per month). The expectation for the beginning of 2007 is 30 – 46 fewer reports per month (95% confidence interval), but obviously this will ease-off very quickly. As a percentage of all processed reports, the number of reports referring to content which is “Not Accessible” stood at approximately 0.20% at the last quarter of 2006. The maximum achieved in any one month was 13%. Between September 2004 and December 2006 it decreased fast at an average rate of 0.12 percent units per month (equivalently, -61% per month). The expectation for the beginning of 2007 is 0.11 – 0.14 fewer percent units per month (95% confidence interval) but obviously this will ease-off right away.
Page 43 of 99
Not for publication without permission © INHOPE 2007
14% 14%
8,000 7,000 6,000 5,000 4,000 3,000 2,000 1,000 0 04-09 04-12 05-03 05-06
12% 12% 10% 10% 8% 8% 6% 6% 4% 4% 2% 2% 0% 0% 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03
05-09
05-12
06-03
06-06
06-09
06-12
07-03
Number of reports referring to content Not Accessible and as a percentage of the total reports processed (inset). • • • • The actual measurements are represented by the points joined with the dotted line (all in red). The smooth curve (in blue) represents the estimated trend line. The straight line (in green) represents the linear trend whose slope gives the average monthly rate of change, quoted above. See also “NOTES” in p. 13.
INHOPE COMMENT The number of reports “Not Accessible” is an insignificant portion of the total reports processed (0.20%) and in addition they ‘suffer’ from a dramatic negative growth rate (-61% per month). This implies that according to current trends, there will not be any “Not Accessible” reports very soon. A similar, but not so dramatic picture for the reports Not Found may also be drawn: They are 1% of the total reports processed and they ‘suffer’ from a negative growth rate (-9% per month). If this negative growth rate is sustained, this category’s lifetime is 11 months. Equally interesting is the fact that the number of Reports Outside the Hotline Remit count for about 7% of all reports processed and growing at +43% per year. Soon some hotlines will have to find ways to avoid having to receive such reports, or change their sphere of activity.
Page 44 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
2.6
CLASSIFICATION BY INTERNET SERVICES
OVERVIEW “Internet Services” indicate the type of Internet service ‘responsible’ for the report/query. Each received report is also classified according to the Internet Service about which a complaint is reported. These services may be: Websites, Email/Spam, Usenet, Chat /real-time, Peer-to-Peer (P2P), FTP (file transfer protocol), Instant Messaging /streaming, Mobile / WAP Services, etc. TREND As depicted in the pie-chart below, 59% of the reports received concern Websites, 30% concern Email/Spam, 1.2% concern Chat/Messaging, 0.8% concern P2P and 9% concern the remaining services (Other).
59%
30%
9%
1.2% 0.8%
Web
Email/Spam
Chat/Messaging
P2P
Other
Chart of different Internet Services, and their contribution to the Total number of Reports Received (trend average for last quarter of 2006).
Page 45 of 99
Not for publication without permission © INHOPE 2007
TREND The number of “Received Reports about Websites” stood at approximately 21,000 reports per month, at the last quarter of 2006. The maximum received in any one month was 31,700 reports. The average rate of increase, between September 2004 and December 2006, was 320 additional reports per month (equivalently, +18% per year). The expectation for the beginning of 2007 is 190 – 440 additional reports per month (95% confidence interval). As a percentage of all received reports, the number of “Received Reports about Websites” stood at approximately 59% at the last quarter of 2006. The maximum achieved in any one month was 70.7%. There was no statistically significant monthly change, between September 2004 and December 2006 and none is expected for the beginning of 2007.
80% 80% 70% 70% 60% 60% 50% 50% 40% 40% 30% 30% 20% 20% 10% 10% 0% 0% 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03
35,000 30,000 25,000 20,000 15,000 10,000 5,000 0 04-09
04-12 05-03
05-06 05-09
05-12
06-03 06-06
06-09 06-12
07-03
Number of Received Reports about Websites
and as a percentage of all received reports (inset)
The
number
of
“Received
Reports
about
Email/Spam”
stood
at
approximately 11,000 reports per month, at the last quarter of 2006. The maximum received in any one month was 41,000 reports. The average rate of increase, between September 2004 and December 2006, was 140 additional reports per month (equivalently, +16% per year). The expectation for the beginning of 2007 is 65 – 210 additional reports per month (95% confidence interval).
Page 46 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
As a percentage of all received reports, the number of “Received Reports about Email/Spam” stood at approximately 30% at the last quarter of 2006. The maximum achieved in any one month was 54.6%. Between September 2004 and December 2006 it decreased slowly at an average rate of 0.10 percent units per month (equivalently, -4.0% per year). The expectation for the beginning of 2007 is 0.002 – 0.200 fewer percent units per month (95% confidence interval).
45.000 40.000 35.000 30.000 25.000 20.000 15.000 10.000 5.000 0 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03
60% 60% 50% 50% 40% 40% 30% 30% 20% 20% 10% 10% 0% 0% 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03
Number of Received Reports about Email/Spam
and as a percentage of all received reports (inset) • • • • The actual measurements are represented by the points joined with the dotted line (all in red). The smooth curve (in blue) represents the estimated trend line. The straight line (in green) represents the linear trend whose slope gives the average monthly rate of change, quoted above. See also “NOTES” in p. 13.
INHOPE COMMENT Reports related to Usenet newsgroups are included in the chart on p45 chart as “other Internet Services”. These reports are primarily generated from hotline “proactive” activity and not from reports received from the public. Some hotlines, in agreement with their national stakeholders, proactively search publicly accessible Usenet newsgroups for illegal content using specially developed software tools. The increase in use of social networking websites as part of Web 2.0 on the Internet increased dramatically during 2007 and does not register as significant for the period covered by this report.
Page 47 of 99
Not for publication without permission © INHOPE 2007
Websites and Email/Spam currently account for nearly 90% of all reports received and their share have remained constant over the last two years. The change rates of the other services are not dramatic, so the shares of the various Internet services are expected to remain stable.
Page 48 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
3
SPECIFIC INTERNET TRENDS
OVERVIEW This section describes the trends relating to specific categories of content considered to be illegal or harmful in at least one INHOPE country. The section focuses exclusively on processed (hotline-determined) reports: 1. Child Pornography 2. Other Child-Related Content 3. Racism & Xenophobia 4. Extreme Adult Content 5. Adult Pornography 6. Other Illegal Content
19% 50% 28%
0.7%
Child Porn Other Adult Porn
0.8% 1.1%
Extreme Racism Other illegal
Categories, and their contribution to the Total number of Illegal or Harmful Reports Processed (average of trend at the last quarter of 2006)
Page 49 of 99
Not for publication without permission © INHOPE 2007
3.1 CHILD PORNOGRAPHY
OVERVIEW “Child Pornography” are reports determined by the hotlines to refer to content which is considered as child pornography under national law. The Council of Europe definition of child pornography in the Cybercrime Convention states that the term "child pornography" shall include pornographic material that visually depicts: a) a minor engaged in sexually explicit conduct; b) a person appearing to be a minor engaged in sexually explicit conduct; c) realistic images representing a minor engaged in sexually explicit conduct. This is, by far, the most important type of content given its appalling nature, the large numbers of such reports and, if processed speedily, the potential to lead to the rescue of a child in danger. Child Pornography as a percentage of the Total Processed Reports about Illegal or Harmful Content14 is a very useful statistic because it gives the relative weight of this type of reported content within the illegal or harmful sectors of the Internet, as charted by the INHOPE network. TREND The number of “Processed Reports about Child Pornography” stood at approximately 9,600 reports per month, at the last quarter of 2006. The maximum processed in any one month was 12,100 reports. The average rate of increase, between September 2004 and December 2006, was 120 additional reports per month (equivalently, +15% per year). The expectation for the beginning of 2007 is 85 – 150 additional reports per month (95% confidence interval).
14
Illegal or harmful Internet content, as a whole, was discussed in Section 2.3
Not for publication without permission
Page 50 of 99 © INHOPE 2007
September 2007
Trend Report
14,000 12,000 10,000 8,000 6,000 4,000 2,000 0 04-09
04-12 05-03
05-06 05-09
05-12
06-03 06-06
06-09 06-12
07-03
Number of Processed Reports about Child Pornography
TREND As a percentage of Total Processed Reports about Illegal or Harmful Content, “Child Pornography” stood at approximately 50% at the last quarter of 2006. The maximum achieved in any one month was 62.7%. Between September 2004 and December 2006 it decreased slowly at an average rate of 0.16 percent units per month (equivalently, -3.9% per year). The expectation for the beginning of 2007 is 0.10 – 0.23 fewer percent units per month (95% confidence interval).
70% 60% 50% 40% 30% 20% 10% 0% 04-09
04-12
05-03
05-06
05-09
05-12
06-03
06-06
06-09
06-12
07-03
Child Pornography as a percentage of Total Processed Reports about Illegal or Harmful Content
• • • •
The actual measurements are represented by the points joined with the dotted line (all in red). The smooth curve (in blue) represents the estimated trend line. The straight line (in green) represents the linear trend whose slope gives the average monthly rate of change, quoted above. See also “NOTES” in p. 13.
Not for publication without permission © INHOPE 2007
Page 51 of 99
INHOPE COMMENT Half of the total processed “Reports about Illegal or Harmful Content” concern Child Pornography. This share has remained almost stable over the last two years (it looses 2 percent points per year). On the other hand, Child Pornography reports in absolute numbers have increased by 15% per year, over the past two years, to reach almost 10,000 reports per month.
Page 52 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
3.2
OTHER CHILD RELATED CONTENT
OVERVIEW “Other Child Related Content” is a category which refers to other (than Child Pornography) illegal or harmful content relating to children. It comprises the following INHOPE template categories: Child Trafficking, Child Sex Tourism, Child Nudism, Child Grooming Activities, Child Erotica/Inappropriate Images of Children and Adult Pornography Accessible to Children. It is a very useful statistic (both on its own but also in combination with Child Pornography) because it gives a measure of child-related concerns, within the illegal and harmful content of the Internet. Some of this content is considered illegal in most countries but other (such as nude pictures of children on a beach) is not considered illegal in most countries. These latter images can give rise to concern when they are found as part of a large collection of other illegal images of children. TREND The number of “Processed Reports about Other Child Related Content” stood at approximately 3,600 reports per month, at the last quarter of 2006. The maximum processed in any one month was 10,000 reports. The average rate of increase, between September 2004 and December 2006, was 51 additional reports per month (equivalently, +17% per year). The expectation for the beginning of 2007 is 31 – 71 additional reports per month (95% confidence interval). As a percentage of Total Processed Reports about Illegal or Harmful Content, “Other Child Related Content” stood at approximately 19% at the last quarter of 2006. The maximum achieved in any one month was 40.8%.
Page 53 of 99 Not for publication without permission © INHOPE 2007
There was no statistically significant monthly change, between September 2004 and December 2006 and none is expected for the beginning of 2007.
70% 70% 60% 60%
12,000 10,000 8,000 6,000 4,000 2,000 0 04-09
50% 50% 40% 40% 30% 30% 20% 20% 10% 10% 0% 0% 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03
04-12 05-03
05-06 05-09
05-12
06-03 06-06
06-09 06-12
07-03
Number of Processed Reports about Other Child-Related Content and as a percentage of the Total Processed Reports about Illegal or Harmful Content (inset)
• • • •
The actual measurements are represented by the points joined with the dotted line (all in red). The smooth curve (in blue) represents the estimated trend line. The straight line (in green) represents the linear trend whose slope gives the average monthly rate of change, quoted above. See also “NOTES” in p. 13.
INHOPE COMMENT Other Child-Related Content accounted for 19% of Total Processed Reports about Illegal or Harmful Content. This share has remained stable over the last two years. Therefore we know that 7 out of every 10 processed reports about illegal or harmful content refer to Child Pornography, or Other Child-Related Content.
Page 54 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
3.3
RACISM AND XENOPHOBIA
OVERVIEW “Racism and Xenophobia” are the reports (external, derived, or proactive) determined by the hotlines to refer to content which is defined as racism or xenophobia under national law. The Council of Europe definition states that "racist and xenophobic material" means any written material, any image or any other representation of ideas or theories, which advocates, promotes or incites hatred, discrimination or violence, against any individual or group of individuals, based on race, colour, descent or national or ethnic origin, as well as religion if used as a pretext for any of these factors. This is an important type of content, especially in such a vast multicultural Union like the EU, but also in many other parts of the world. TREND The number of “Processed Reports about Racism and Xenophobia” stood at approximately 220 reports per month, at the last quarter of 2006. The maximum processed in any one month was 340 reports. The average rate of increase, between September 2004 and December 2006, was 5.9 additional reports per month (equivalently, +33% per year). The expectation for the beginning of 2007 is 4.0 – 7.8 additional reports per month (95% confidence interval). As a percentage of Total Processed Reports about Illegal or Harmful Content, “Racism and Xenophobia” stood at approximately 1.1% at the last quarter of 2006. The maximum achieved in any one month was 1.71%. Between September 2004 and December 2006 it increased at an average rate of 0.023 percent units per month (equivalently, +24% per year). The
Page 55 of 99 Not for publication without permission © INHOPE 2007
expectation for the beginning of 2007 is 0.021 – 0.024 additional percent units per month (95% confidence interval).
2.0% 2.0%
400 350 300 250 200 150 100 50
1.5% 1.5% 1.0% 1.0% 0.5% 0.5% 0.0% 0.0% 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03
0 04-09
04-12
05-03
05-06
05-09
05-12
06-03
06-06
06-09
06-12
07-03
Number of Processed Reports about Racism and Xenophobia and as a percentage of the Total Processed Reports about Illegal or Harmful Content (inset)
• • • •
The actual measurements are represented by the points joined with the dotted line (all in red). The smooth curve (in blue) represents the estimated trend line. The straight line (in green) represents the linear trend whose slope gives the average monthly rate of change, quoted above. See also “NOTES” in p. 13.
INHOPE COMMENT Reports on Racism and Xenophobia stand at about 340 per month, occupying only 1% of the Total Processed Reports about Illegal or Harmful Content, but their share grows at +24% per year.
Page 56 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
3.4
EXTREME ADULT CONTENT
OVERVIEW “Extreme Adult Content”, is defined to be content which includes extreme sexual and physical violence, non-consensual sexual acts and other types of pornographic content deemed illegal under national law. For example, rape websites etc. TREND The number of “Processed Reports about Extreme Adult Content” stood at approximately 150 reports per month, at the last quarter of 2006. The maximum processed in any one month was 1,010 reports. The average rate of decrease, between September 2004 and December 2006, was 5.2 fewer reports per month (equivalently, -43% per year). The expectation for the beginning of 2007 is 2.2 – 8.3 fewer reports per month (95% confidence interval). As a percentage of Total Processed Reports about Illegal or Harmful Content, “Extreme Adult Content” stood at approximately 0.77% at the last quarter of 2006. The maximum achieved in any one month was 5.59%. Between September 2004 and December 2006 it decreased at an average rate of 0.061 percent units per month (equivalently, -94% per year). The expectation for the beginning of 2007 is 0.054 – 0.068 fewer percent units per month (95% confidence interval), but this is expected to ease-off soon.
Page 57 of 99
Not for publication without permission © INHOPE 2007
1,200 1,000 800 600 400 200 0 04-09
6% 6% 5% 5% 4% 4% 3% 3% 2% 2% 1% 1% 0% 0% 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03
04-12
05-03
05-06
05-09
05-12
06-03
06-06
06-09
06-12
07-03
Number of Processed Reports about Extreme Adult Content and as a percentage of the Total Processed Reports about Illegal or Harmful Content (inset)
• • • •
The actual measurements are represented by the points joined with the dotted line (all in red). The smooth curve (in blue) represents the estimated trend line. The straight line (in green) represents the linear trend whose slope gives the average monthly rate of change, quoted above. See also “NOTES” in p. 13.
INHOPE COMMENT Extreme Adult Content has a low share of the Total Processed Reports about Illegal or Harmful Content (less than 1%) and more remarkably it is loosing ground at a fast rate of 8% per month. At this rate it will disappear very soon.
Page 58 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
3.5
ADULT PORNOGRAPHY
OVERVIEW “Adult Pornography” is pornographic content which in most countries does not contravene national law. Its statistics are significant, not only because of its numbers but also because there is occasional confusion with Child Pornography when the borderline between legal and illegal is unclear. TREND The number of “Processed Reports about Adult Pornography” stood at approximately 5,300 reports per month, at the last quarter of 2006. The maximum processed in any one month was 8,610 reports. The average rate of increase, between September 2004 and December 2006, was 100 additional reports per month (equivalently, +24% per year). The expectation for the beginning of 2007 is 80 – 130 additional reports per month (95% confidence interval). As a percentage of Total Processed Reports about Illegal or Harmful Content, “Adult Pornography” stood at approximately 28% at the last quarter of 2006. The maximum achieved in any one month was 42.1%. Between September 2004 and December 2006 it increased slowly at an average rate of 0.21 percent units per month (equivalently, +8.9% per year). The expectation for the beginning of 2007 is 0.17 – 0.24 additional percent units per month (95% confidence interval).
Page 59 of 99
Not for publication without permission © INHOPE 2007
50% 50% 45% 45% 40% 40% 35% 35% 30% 30% 25% 25%
10,000 9,000 8,000 7,000 6,000 5,000 4,000 3,000 2,000 1,000 0 04-09 04-12 05-03 05-06 05-09
20% 20% 15% 15% 10% 10% 5% 5% 0% 0% 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03
05-12
06-03 06-06
06-09 06-12
07-03
Number of Processed Reports about Adult Pornography and as a percentage of the Total Processed Reports about Illegal or Harmful Content (inset)
• • • •
The actual measurements are represented by the points joined with the dotted line (all in red). The smooth curve (in blue) represents the estimated trend line. The straight line (in green) represents the linear trend whose slope gives the average monthly rate of change, quoted above. See also “NOTES” in p. 13.
INHOPE COMMENT Adult Pornography occupies 28% of Total Processed Reports about Illegal or Harmful Content and increases its share at a slow rate of 9% per year.
Page 60 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
3.6
OTHER ILLEGAL CONTENT
OVERVIEW “Other Illegal Content”, comprises the following INHOPE template
categories: Promoting Violence against an Individual, Terrorism and Drugs. TREND The number of “Processed Reports about Other Illegal Content” stood at approximately 130 reports per month, at the last quarter of 2006. The maximum processed in any one month was 1,230 reports. The average rate of increase, between September 2004 and December 2006, was 4.8 additional reports per month (equivalently, +43% per year). The expectation for the beginning of 2007 is 3.0 – 6.6 additional reports per month (95% confidence interval). As a percentage of Total Processed Reports about Illegal or Harmful Content, “Other Illegal Content” stood at approximately 0.70% at the last quarter of 2006. The maximum achieved in any one month was 6.15%. Between September 2004 and December 2006 it increased at an average rate of 0.024 percent units per month (equivalently, +41% per year). The expectation for the beginning of 2007 is 0.023 – 0.025 additional percent units per month (95% confidence interval).
Page 61 of 99
Not for publication without permission © INHOPE 2007
1,400 1,200 1,000 800 600 400 200 0 04-09
7% 7% 6% 6% 5% 5% 4% 4% 3% 3% 2% 2% 1% 1% 0% 0% 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03
04-12
05-03
05-06
05-09
05-12
06-03
06-06
06-09
06-12
07-03
Number of Processed Reports of “Other Illegal Content and as a percentage of the Total Processed Reports about Illegal or Harmful Content (inset)
• • • •
The actual measurements are represented by the points joined with the dotted line (all in red). The smooth curve (in blue) represents the estimated trend line. The straight line (in green) represents the linear trend whose slope gives the average monthly rate of change, quoted above. See also “NOTES” in p. 13.
INHOPE COMMENT Other Illegal Content concerns 130 reports per month, increasing at 43% per year and contains categories, whose growth will be important to monitor.
Page 62 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
4 WHAT HAPPENS TO REPORTS ONCE THEY HAVE BEEN PROCESSED BY A HOTLINE?
OVERVIEW This section describes the actions which the INHOPE network takes in relation to reports which have been processed by the hotline according to national law. Mostly, illegal content is reported to agreed contacts inside the most appropriate law enforcement agency but sometimes, with advance agreement, other actions can be taken. • Total Resulting Hotline Actions
o
Reports Transmitted to Law Enforcement Reports Transmitted to Hosting ISP Reports Forwarded to another INHOPE Hotline Other Resulting Hotline Actions
o o o
72% 14% 11% 3%
Police
ISP
Hotline
Other
Actions, and their contribution to the Total Resulting Hotline Actions (average of trend at the last quarter of 2006)
Page 63 of 99
Not for publication without permission © INHOPE 2007
4.1
TOTAL RESULTING HOTLINE ACTIONS
OVERVIEW “Total Resulting Hotline Actions” are the actions taken by hotlines in relation to the processed reports which require further action by the hotline. This may include forwarding reports to Law Enforcement, to Hosting Provider, etc. These statistics reflect the action workload of the hotline after a report has been processed. This workload is in general under-estimated by the number of “Unique Actionable Reports” as often one report may have to be forwarded to multiple authorities. TREND The number of “Total Resulting Hotline Actions” stood at approximately 46,000 actions per month, at the last quarter of 2006. The maximum completed in any one month was 209,000 actions. The average rate of increase, between September 2004 and December 2006, was 790 additional actions per month (equivalently, +21% per year). The expectation for the beginning of 2007 is 660 – 920 additional actions per month (95% confidence interval).
Page 64 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
250,000 200,000 150,000 100,000 50,000 0 04-09
04-12
05-03
05-06
05-09
05-12 06-03
06-06
06-09
06-12
07-03
Number of Total Resulting Hotline Actions • • • • The actual measurements are represented by the points joined with the dotted line (all in red). The smooth curve (in blue) represents the estimated trend line. The straight line (in green) represents the linear trend whose slope gives the average monthly rate of change, quoted above. See also “NOTES” in p. 13.
INHOPE COMMENT In some countries a confirmed report must be sent to the police even if it has already been forwarded to another Hotline. Therefore, this one report will result in “Transmission to Police” being incremented by 1 and “Forward to other INHOPE Hotline” also being incremented by 1.
Page 65 of 99
Not for publication without permission © INHOPE 2007
4.1.1
REPORTS TRANSMITTED TO LAW ENFORCEMENT
OVERVIEW “Reports Transmitted to Law-Enforcement” are the reports forwarded by the hotlines to Police, because the hotlines believe that action should be taken by the Police. TREND The number of “Reports Transmitted to Law-Enforcement” stood at approximately 6,400 reports per month, at the last quarter of 2006. The maximum transmitted in any one month was 7,830 reports. The average rate of increase, between September 2004 and December 2006, was 52 additional reports per month (equivalently, +9.8% per year). The expectation for the beginning of 2007 is 35 – 70 additional reports per month (95% confidence interval). As a percentage of Total Resulting Hotline Actions, the number of “Reports Transmitted to Law-Enforcement” stood at approximately 14% at the last quarter of 2006. The maximum achieved in any one month was 57.1%. Between September 2004 and December 2006 it decreased at an average rate of 0.23 percent units per month (equivalently, -20% per year). The expectation for the beginning of 2007 is 0.14 – 0.32 fewer percent units per month (95% confidence interval).
Page 66 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
8,000 7,000 6,000 5,000 4,000 3,000 2,000 1,000 0 04-09 04-12 05-03 05-06 05-09
60% 60% 50% 50% 40% 40% 30% 30% 20% 20% 10% 10% 0% 0% 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03
05-12 06-03 06-06 06-09
06-12 07-03
Number of reports Transmitted to Law Enforcement and as a percentage of the Total Resulting Hotline Actions (inset)
• • • •
The actual measurements are represented by the points joined with the dotted line (all in red). The smooth curve (in blue) represents the estimated trend line. The straight line (in green) represents the linear trend whose slope gives the average monthly rate of change, quoted above. See also “NOTES” in p. 13.
INHOPE COMMENT This is an important statistic, as it gives a clear synopsis of the co-operation between the Police and the hotlines. In the period covered by this report over 162,000 reports were transmitted to law enforcement. These reports relate to content which is considered illegal under national law. INHOPE hotlines rarely receive structured feedback from law enforcement in relation to these reports. This is due to the complexity of online transnational crime investigations, the time delay between receiving the report and completing a successful investigation, the limitations of national legislation on the issue of giving feedback, the challenge of relating specific reports received from an INHOPE hotline to a specific criminal activity and perhaps other reasons. This lack of feedback can be quite demotivating for hotline staff who labour each day looking at harmful and illegal content in order to report these for investigation. INHOPE hotlines are working to adopt signed Memorandum of Understanding with national law enforcement which would include feedback in relation to reports forwarded for investigation.
Page 67 of 99 Not for publication without permission © INHOPE 2007
In addition some national law enforcement agencies request the local hotline not to forward reports to them which relate to content outside their jurisdiction in locations where an INHOPE hotline exists since they have confidence in INHOPE network system of exchanging reports. Other agencies request hotlines to send all reports relating to illegal content to them so that they can ensure it is sent through international police channels for investigation. Also they can then keep track of the trends of illegal activity on the Internet. At an international level, INHOPE is an official observer on the Interpol Crimes Against Children subgroup. Since its start in 1999, INHOPE and Europol, following the extension of their mandate since January 2000, have shared common aims to tackle the growing problem of Child Pornography on the Internet. INHOPE gave a presentation on its activities at the 2005 Europol training for law enforcement and the judiciary combating the sexual exploitation of children on the Internet. Since 2000 this annual Europol training enhances the knowledge of detecting, investigating these crimes in view of the identification and saving of the abused victims and the prosecution of the offenders. In addition, INHOPE has given presentation on its activities at Europol's annual Child Abuse Expert meeting. Europol have also addressed the INHOPE hotline network at the INHOPE's members meeting explaining their role and discussing current trends related to the sexual exploitation of children on the Internet. More information on Europol's approach to combating Child Pornography can be found by accessing their Child Abuse Fact Sheet.
Page 68 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
4.1.2 REPORTS TRANSMITTED TO HOSTING ISP
OVERVIEW “Reports Transmitted to Hosting ISP” are the reports forwarded by the hotlines to the hosting Internet Service Provider, because the hotlines believe that urgent action should be taken by the Internet Service Provider. This action could include removing the offending content, reporting the offending content to law enforcement, preserving the log files relating to the offending content for use later, reporting the abuse to the customer who placed the content online, etc. TREND The number of “Reports Transmitted to Hosting ISP” stood at approximately 5,000 reports per month, at the last quarter of 2006. The maximum transmitted in any one month was 36,400 reports. The average rate of increase, between September 2004 and December 2006, was 96 additional reports per month (equivalently, +23% per year). The expectation for the beginning of 2007 is 23 – 170 additional reports per month (95% confidence interval). As a percentage of Total Resulting Hotline Actions, the number of “Reports Transmitted to Hosting ISP” stood at approximately 11% at the last quarter of 2006. The maximum achieved in any one month was 23.3%. There was no statistically significant monthly change, between September 2004 and December 2006 and none is expected for the beginning of 2007.
Page 69 of 99
Not for publication without permission © INHOPE 2007
40,000 35,000 30,000 25,000 20,000 15,000 10,000 5,000
30% 30% 25% 25% 20% 20% 15% 15% 10% 10% 5% 5% 0% 0% 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03
0 04-09
04-12 05-03
05-06 05-09
05-12
06-03 06-06
06-09 06-12
07-03
Number of reports Transmitted to Hosting ISP and as a percentage of the Total Resulting Hotline Actions (inset)
• • • •
The actual measurements are represented by the points joined with the dotted line (all in red). The smooth curve (in blue) represents the estimated trend line. The straight line (in green) represents the linear trend whose slope gives the average monthly rate of change, quoted above. See also “NOTES” in p. 13.
INHOPE COMMENT Reporting to the hosting Internet Service Provider and the content owner is a step which needs to be taken carefully in order not to interfere with any ongoing law enforcement investigation. However, sometimes with extremely volatile content – where there is ongoing immediate abuse of a child immediate action is required to ensure that the child‘s safety is paramount. Depending on national legislation, the ISP sometimes prefers not to be informed about potentially illegal content.
Page 70 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
4.1.3 REPORTS FORWARDED TO ANOTHER INHOPE HOTLINE
OVERVIEW “Reports Forwarded to another INHOPE Hotline” are the reports which are forwarded by the hotlines to another INHOPE hotline, because the Internet content was traced to the jurisdiction of that hotline. TREND The number of “Reports Forwarded to another INHOPE Hotline” stood at approximately 1,400 reports per month, at the last quarter of 2006. The maximum forwarded in any one month was 2,020 reports. The average rate of increase, between September 2004 and December 2006, was 16 additional reports per month (equivalently, +13% per year). The expectation for the beginning of 2007 is 10 – 21 additional reports per month (95% confidence interval). As a percentage of Total Resulting Hotline Actions, the number of “Reports Forwarded to another INHOPE Hotline” stood at approximately 3.0% at the last quarter of 2006. The maximum achieved in any one month was 10.7%. Between September 2004 and December 2006 it decreased at an average rate of 0.027 percent units per month (equivalently, -11% per year). The expectation for the beginning of 2007 is 0.011 – 0.043 fewer percent units per month (95% confidence interval).
Page 71 of 99
Not for publication without permission © INHOPE 2007
12% 12% 10% 10% 8% 8% 6% 6%
2,500 2,000 1,500 1,000 500 0 04-09
4% 4% 2% 2% 0% 0% 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03
04-12
05-03
05-06
05-09
05-12
06-03
06-06
06-09
06-12
07-03
Number of reports Forwarded to another INHOPE Hotline and as a percentage of the Total Resulting Hotline Actions (inset)
• • • •
The actual measurements are represented by the points joined with the dotted line (all in red). The smooth curve (in blue) represents the estimated trend line. The straight line (in green) represents the linear trend whose slope gives the average monthly rate of change, quoted above. See also “NOTES” in p. 13.
INHOPE COMMENT Once the receiving hotline reviews the reported content according to national law and agreed hotline procedures and determines that the reported content is likely to be illegal, it traces the content nationally and takes action.
Page 72 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
4.1.4 OTHER RESULTING ACTIONS
OVERVIEW “Other Resulting Actions” refer to all the other types of actions, taken by an INHOPE hotline, apart from those covered previously. This comprises to the following actions from the to INHOPE other template: Competent
Transmission
Content
Owner,
Transmission
Authorities / Bodies and Follow-up Actions / Response to Reporter. TREND The number of “Other Resulting Actions” stood at approximately 34,000 actions per month, at the last quarter of 2006. The maximum performed in any one month was 165,000 actions. The average rate of increase, between September 2004 and December 2006, was 630 additional actions per month (equivalently, +22% per year). The expectation for the beginning of 2007 is 360 – 900 additional actions per month (95% confidence interval). As a percentage of Total Resulting Hotline Actions, the number of “Other Resulting Actions” stood at approximately 72% at the last quarter of 2006. The maximum achieved in any one month was 83.2%. Between September 2004 and December 2006 it increased slowly at an average rate of 0.23 percent units per month (equivalently, +3.8% per year). The expectation for the beginning of 2007 is 0.09 – 0.37 additional percent units per month (95% confidence interval).
Page 73 of 99
Not for publication without permission © INHOPE 2007
100% 100% 90% 90% 80% 80% 70% 70% 60% 60% 50% 50% 40% 40% 30% 30% 20% 20% 10% 10% 0% 0% 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03
180,000 160,000 140,000 120,000 100,000 80,000 60,000 40,000 20,000
0 04-09 04-12 05-03 05-06 05-09 05-12 06-03 06-06 06-09 06-12 07-03
Number of Other Resulting Actions and as a percentage of the Total Resulting Hotline Actions (inset)
• • • •
The actual measurements are represented by the points joined with the dotted line (all in red). The smooth curve (in blue) represents the estimated trend line. The straight line (in green) represents the linear trend whose slope gives the average monthly rate of change, quoted above. See also “NOTES” in p. 13.
INHOPE COMMENT Other Resulting Actions is a sizeable category of 34,000 per month increasing at +22% per year and occupying 72% (and growing) of the Total Resulting Hotline Actions. In the future, it will be important to divide this into sub-categories.
Page 74 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
5.
ADVANCED TREND RESULTS
OVERVIEW This section describes some complex trend comparisons which has been included for the advanced reader only. This section compares the relative significance of received versus processed content, for some categories of interest, like Child Pornography, Adult Pornography, etc. It has been observed that the share of, for example, Child Pornography within the Illegal or Harmful Content, is different for the received (by the general public) reports than it is for the processed (hotline-determined) reports.
For example, at the last quarter of 2006 the average of the trend for Child Pornography, as a percent of the Total Illegal or Harmful Content, was 60% for the received reports against 50% for the processed reports. The ratio of the two is 60/50 = 120% and is a measure of the overestimation of the particular category of content (against other illegal or harmful content) by the general public. Obviously, when the comparison is less than 100% we have underestimation.
In what follows, Comparison Ratio (REC%/PRO%) will imply the ratio of the percent for the received, over the percent for the processed.
CHILD PORNOGRAPHY ON THE INTERNET - TREND The “Comparison Ratio (REC%/PRO%) for Child Pornography” stood at approximately 120% at the last quarter of 2006. The maximum reached in any one month was 160%. The average rate of decrease, between
Page 75 of 99 Not for publication without permission © INHOPE 2007
September 2004 and December 2006, was 8 percent units per year (-7% per year).
200%
150%
100%
50%
0% 04-09 04-12 05-03
05-06 05-09 05-12 06-03 06-06
06-09 06-12 07-03
Comparison Ratio (REC%/PRO%) for Child Pornography
OTHER CHILD-RELATED CONTENT15 ON THE INTERNET - TREND The “Comparison Ratio (REC%/PRO%) for Other Child-related Content” stood at approximately 59% at the last quarter of 2006. The maximum reached in any one month was 77%. The average rate of increase, between September 2004 and December 2006, was 7 percent units per year (+12% per year).
200%
150%
100%
50%
0% 04-09 04-12 05-03
05-06 05-09 05-12 06-03 06-06
06-09 06-12 07-03
Comparison Ratio (REC%/PRO%) for Other Child-related Content
This comprises the following INHOPE template categories: Child Trafficking, Child Sex Tourism, Child Nudism, Child Grooming Activities, Child Erotica/Inappropriate Images of Children and Adult Pornography Accessible to Children.
Page 76 of 99 © INHOPE 2007
15
Not for publication without permission
September 2007
Trend Report
CHILD-RELATED CONTENT16 ON THE INTERNET - TREND The “Comparison Ratio (REC%/PRO%) for Child-related Content” stood at approximately 100% at the last quarter of 2006. The maximum reached in any one month was 122%. The average rate of change, between September 2004 and December 2006, was practically nil.
200%
150%
100%
50%
0% 04-09 04-12 05-03
05-06 05-09 05-12 06-03 06-06
06-09 06-12 07-03
Comparison Ratio (REC%/PRO%) for Child-related Content
ADULT PORNOGRAPHY ON THE INTERNET - TREND The “Comparison Ratio (REC%/PRO%) for Adult Pornography” stood at approximately 92% at the last quarter of 2006. The maximum reached in any one month was 260%. The average rate of increase, between September 2004 and December 2006, was 11 percent units per year (+12% per year).
200%
150%
100%
50%
0% 04-09 04-12 05-03
05-06 05-09 05-12 06-03 06-06
06-09 06-12 07-03
Comparison Ratio (REC%/PRO%) for Adult Pornography
16
This comprises Child Pornography and Other Child-related Content (defined above).
Not for publication without permission © INHOPE 2007
Page 77 of 99
EXTREME ADULT CONTENT ON THE INTERNET - TREND The “Comparison Ratio (REC%/PRO%) for Extreme Adult Content” stood at approximately 100% at the last quarter of 2006. The maximum reached in any one month was 150%. The average rate of increase, between September 2004 and December 2006, was 6 percent units per year (+6% per year).
200%
150%
100%
50%
0% 04-09 04-12 05-03
05-06 05-09 05-12 06-03 06-06
06-09 06-12 07-03
Comparison Ratio (REC%/PRO%) for Extreme Adult Content
PORNOGRAPHY17 ON THE INTERNET - TREND The “Comparison Ratio (REC%/PRO%) for Pornography” stood at
approximately 93% at the last quarter of 2006. The maximum reached in any one month was 240%. The average rate of increase, between September 2004 and December 2006, was 11.0 percent units per year (+12% per year).
200%
150%
100%
50%
0% 04-09 04-12 05-03
05-06 05-09 05-12 06-03 06-06
06-09 06-12 07-03
Comparison Ratio (REC%/PRO%) for Pornography
17
This comprises Adult Pornography and Extreme Adult Content.
Not for publication without permission
Page 78 of 99 © INHOPE 2007
September 2007
Trend Report
CHILD-RELATED CONTENT AND PORNOGRAPHY ON THE INTERNET - TREND The “Comparison Ratio (REC%/PRO%) for Child-related Content and Pornography” stood at approximately 99% at the last quarter of 2006. The maximum reached in any one month was 103%. The average rate of change, between September 2004 and December 2006, was practically nil.
200%
150%
100%
50%
0% 04-09 04-12 05-03
05-06 05-09 05-12 06-03 06-06
06-09 06-12 07-03
Comparison Ratio (REC%/PRO%) for Child-related Content and Pornography
• • • •
The actual measurements are represented by the points joined with the dotted line (all in red). The smooth curve (in blue) represents the estimated trend line. The straight line (in green) represents the linear trend whose slope gives the average monthly rate of change, quoted above. See also “NOTES” in p. 13.
INHOPE COMMENT Over the period examined, the above results indicate the following: 1. Child Pornography was overestimated18 by the general public by between 20-40%. This tendency though is definitely decreasing at an average rate of 8 percent units per year. 2. Other Child-related Content was underestimated by the general public by between 40-60%. This tendency though is definitely increasing at an average rate of 7 percent units per year.
Please see Overview above for a complete explanation of the meaning of “over(under)estimation” in this section.
Page 79 of 99 Not for publication without permission © INHOPE 2007
18
3. Child Pornography together with Other Child-related Content was in general correctly estimated by the general public, although there were periods of under- and over-estimation. 4. Adult Pornography was underestimated by the general public by between 10-50%. This tendency though is definitely increasing at an average rate of 11 percent units per year and it is set to switch to overestimation. 5. Extreme Adult Content was underestimated by the general public by between 0-20%. This tendency though is definitely increasing at an average rate of 6 percent units per year. 6. Adult Pornography together with Extreme Adult Content was underestimated by the general public by between 10-50%. This tendency though is definitely increasing at an average rate of 11 percent units per year and it is set to switch to overestimation. 7. All the distinct above categories together were remarkably well estimated by the general public: every single month. The following conclusions may be drawn from the above seven statements: • • Under/over-estimation tends to be corrected with time. The general public tends to overestimate Child Pornography at the expense of Other Child-related Content. • There is no evidence for or against the proposition that Extreme Adult Content is confused with Adult Pornography. • Any under/over estimation by the general public is done at the expense of one of the above considered categories. They summed-up to ~100% for
Page 80 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
CONCLUSION AND FUTURE WORK
CONCLUSION A huge amount of work has been completed to determine the trends highlighted in this first INHOPE trend report. The creation of the statistics Best Practice Paper, the ongoing collection of the individual hotline data, the collation of this data to produce statistics, the processing of these statistics to create trend reports which can withstand critical analysis and the creation of this trend report took a huge effort by many people inside the INHOPE network over the last years. The trends indicate clearly the volume of work which takes place each day within hotlines which are members of the INHOPE International Association of Internet Hotlines. This work is conducted according to internally agreed best practice standards to ensure the highest level of quality possible in the handling of reports of illegal content. The trends indicate that the public in each country where a hotline operates trust their national hotline with a wide range of reports of illegal content. They trust them to take these reports, determine their level of illegality and to forward them to the most appropriate agency across the globe which will take action against the illegal content. In addition, the trends indicate that hotlines play a vital role between the Internet consumer and the major law enforcement agencies to ensure that clearly identified reports of illegal content are sent to the agency with the most capability of taking action against the content and the perpetrators.
Page 81 of 99
Not for publication without permission © INHOPE 2007
The trends sometimes raise more questions than they answer. Often we would like to know the combination of reasons why some trends exist in order to better focus resources and tactics in the future. Such work will be part of the future work plans. FUTURE WORK Cause of trends: We need to analyse trends to determine their cause(s). It is possible to attempt to model the trend of each particular statistic as the weighted sum, or product, of individual causes. If such an attempt is successful, one then may be able to obtain estimates of the values of the individual causes. Comparison of trends: An interesting variation is the comparison of trends between groups of hotlines (geographical and other). This was not done in this first report. A future variation may group together hotlines by • • • • • Age Country social criteria Size of hotline and/or hotline region Language region covered by hotline Etc
The purpose of this additional analysis would be to seek to determine if any common trends could be identified and credited to shared attributes as outlined above.
Page 82 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
APPENDICES
APPENDIX
I
AREAS OF CAUTION EXPLANATIONS FOR VISIBLE TRENDS INHOPE STATISTICS TEMPLATE STATISTICAL PRIMER LIST OF INHOPE MEMBERS
Appendix II Appendix III Appendix IV Appendix V
Page 83 of 99
Not for publication without permission © INHOPE 2007
APPENDIX I AREAS OF CAUTION
DATA CONSISTENCY It is felt that ongoing review of the consistency of the individual hotline contributions is important. This is so because INHOPE consolidates results from different countries, as is the case with 25-30 hotlines and many more hotline administrators. Statistics rules require that experiments are carried out under essentially the same conditions, in other words, all hotline administrators must interpret all the rules in the same way, all the time! So, in order to feel comfortable with the question “Do we count correctly?”, INHOPE must: • Confirm terms19 are defined in a precise and in an easy-tounderstand manner. • Verify that all hotlines interpret the rules in the same manner.
19
used in the INHOPE report template
Not for publication without permission
Page 84 of 99 © INHOPE 2007
September 2007
Trend Report
APPENDIX II EXPLANATIONS FOR VISIBLE TRENDS
A very challenging and most likely unsolvable problem is the separation of the cause factor from the observation factor. For example, a trend increase in the number of reports per year may be due to an increase in illegal activity, or an increase in the hotline visibility, or an increase of public awareness, or an increase in the use of Internet, etc, or on a combination of the above and other factors. When weighing the potential factors that cause a change in a trend, one should take into account national and social ‘particularities’: Has, for example hotline visibility, or Internet penetration, reached saturation levels? Does the general public believe that reporting illegal content is worth doing? Etc. It is obvious that such an attempt to link the observation, which is the INHOPE network reports, with causes is very dangerous if done superficially and by non-experts.
Page 85 of 99
Not for publication without permission © INHOPE 2007
APPENDIX III INHOPE STATISTICS TEMPLATE
Under the INHOPE template system all data is based on the unit of a “report20”. The primary form of “report” is an occurrence of suspected illegal content reported by the public. A report is the input unit of work that a hotline reacts to. In the case of a web-site, it is a reported URL (which may contain multiple legal or illegal images or text, a video or a sound file, etc.). If the person making the report specifically reports separate URLs which are sub-directories of one website they are counted as separate reports.
REPORT TYPES EXTERNAL REPORTS A report is classified as an external report if the report is received by a hotline from a person or an organisation from the wider public. Reports received by a hotline from other INHOPE members are also classified as external reports.
Note: Even if a report is a duplicate, it is counted. (With the exception of reports which are clearly the same report from the same person sent within a short period of time, because e.g. the sender didn’t realise the e-mail had actually been sent and sent it again). External reports may prove to contain no illegal content after investigation by a hotline.
20
Sometimes a report is referred to as a “complaint” Not for publication without permission
Page 86 of 99 © INHOPE 2007
September 2007
Trend Report
DERIVED REPORTS A report is classified as derived if the hotline conducts follow up research from an external report and takes action against different content than that reported as external report.
Example: An external report refers to www.xyz.com but this site has links to www.illegal.com which requires the hotline to open a new report for this “derived” site. In almost all cases a derived report will contain illegal material – because if it does not the hotline should not be recording it.
PROACTIVE REPORTS A report is classified as proactive if the hotline proactively searches for illegal or harmful content on the Internet either manually or through a technical tool and takes action against such content. It should only be recorded as a report when illegal material is found.
REPORT CLASSIFICATIONS TYPES OF CONTENT Reports which are received by a hotline are classified based on the type of content which is reported. Reports can be received about child pornography/child abuse images, racism, extreme adult pornography, etc. The purpose of classification is to indicate the INHOPE view of criminal trends and the type of crimes which are taking place. The information gathered by INHOPE indicates firstly which type of content is being complained about by the Internet public and the public perception of what is illegal on the Internet. The total of these reports is then checked by each hotline and the hotline determination is also recorded in the system. We are therefore able to see the public perception of illegal content and then compare this against the INHOPE expert evaluation of what is illegal content.
Note: There are some hotlines which do not offer the reporting person the option of selecting which type of content is being reported. These
Page 87 of 99 Not for publication without permission © INHOPE 2007
hotlines only accept reports about one type of content e.g. child pornography (child abuse, child abduction, etc) and therefore the hotline assumes that all reports received refer to child pornography. This assumption cannot be confirmed with the public who might have wanted to report a different type of illegal activity and when presented with no option to clarify or specify the type of content, continue to submit the offending content without clarification.
TYPE OF INTERNET SERVICE Reports which are received by a hotline are classified based on the type of Internet service which is complained about. These categories include websites, chat rooms, P2P, newsgroups, etc. The purpose is to indicate to INHOPE whether criminal activities are increasing in websites, P2P, or chat rooms, etc. The INHOPE view is completely dependent on the quality of reports received from the wider Internet public and the range of Internet services they use and may not be an objective indication of the scale of criminal activity in any of these Internet services. Indeed some services such as P2P are very difficult for the public to report illegal activity about and can therefore be easily under-reported.
ACTIONS TAKEN Once a report is processed by the hotline and then confirmed as likely to be illegal, there are a number of steps or actions which a hotline must take. Of course, the first step is to identify the apparent location of the Internet content. The number of actions taken can be greater than the number of actionable records since one report may have to be forwarded to multiple authorities.
Note: In some countries a confirmed report must be sent to the police even if it has already been forwarded to another Hotline. Therefore, that one
Page 88 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
report will result in “Transmission to Police” being incremented by 1 and “Forward to other INHOPE Hotline” also being incremented by 1. If the report also required other follow-up actions, such as response to a reporter (especially if it was another INHOPE Hotline), then this box would also be incremented by 1.
TRANSMISSION TO POLICE This refers to reports which are passed to national Law Enforcement and also include reports where the hotline knows that national Law Enforcement pass it through international police channels (i.e. Interpol) because it is hosted in a different jurisdiction.
TRANSMISSION TO INTERNET SERVICE PROVIDER (ISP) OR HOSTING SERVICE This is used when the hotline passes a report to an ISP or Hosting Service. In some countries, hotlines can issue “Take Down Notices”.
TRANSMISSION TO CONTENT OWNER In some countries, hotlines can notify content owners (e.g. bulletin board) that their facility has been misused, so that the offending content may be removed from public access.
TRANSMISSION TO OTHER COMPETENT AUTHORITIES These are authorities or bodies in the hotline’s home country other than law enforcement (police). In some countries, for example, incidents of racism have to be forwarded to a Government agency in addition to law enforcement.
FORWARD TO OTHER INHOPE HOTLINES A report about content that is probably illegal in the home jurisdiction and apparently located in a country that has an INHOPE Hotline is forwarded to that other INHOPE Hotline. This should be incremented even if you must also notify law enforcement.
Page 89 of 99
Not for publication without permission © INHOPE 2007
FOLLOW-UP ACTIONS / RESPONSE TO REPORTER This only applies to unique actionable reports. This is only for follow-up actions, such as reporting a result back to another Hotline or to a reporter who has requested that the Hotline tell them the result of their report. This includes responses even if they are automated. It does not include the original acknowledgement issued when a report is received which is recorded below.
OTHER HOTLINE ACTIONS These statistics apply to all reports whether actionable or not.
RECEIPT ACKNOWLEDGEMENTS Any form of acknowledgment sent by a hotline on receipt of a report from any external source. This includes a response to inform someone that their report is “outside the Hotline remit”.
EXPLANATORY RESPONSE Any form of response that involves the hotline providing an explanation or advice.
Example: an explanation of legal reasons why the Hotline can not pursue a particular complaint about content that is illegal in their own jurisdiction but has not been removed because the content is legal in the jurisdiction where it is hosted.
Page 90 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
APPENDIX IV STATISTICAL PRIMER
Statistics is a mathematical science that examines ways to process and analyse data. Statistics provide procedures that collect and transform data in ways that are useful to decision makers.
Background Note Historically, the word statistics derives from the New Latin term statisticum collegium ("council of state"). The German Statistik originally designated the analysis of data about the state, signifying the "science of state" (then called political arithmetic in English). It acquired the meaning of the collection and classification of data generally in the early 19th century.21
INTRODUCTION To understand anything about statistics one must first understand the meaning of the following terms: 1. Variables are characteristics of items or individuals
Example: your gender, your country of residence, the amount of money you have in your wallet at this point in time, etc. Variables take on values. For example, your gender takes the values “male” or “female”, etc.
2. A population consists of all the members of a group about which we want to draw a conclusion
Example: All the registered voters of the city of Vienna, as defined on 25 May 2007, at 9 am, etc.
3. A sample is the portion of the population selected for analysis
21
See, for example, Wikipedia.
Not for publication without permission © INHOPE 2007
Page 91 of 99
Example: A group of 1,000 registered voters of the city of Vienna, etc.
4. A parameter is a numerical measure that describes a characteristic of a population
Example: The % of males among the registered voters of the city of Vienna. Let us assume that we know this to be 51.2%.
5. A statistic is a numerical measure that describes a characteristic of a sample
Example: The % of males among the 1,000 strong sample of city voters. Let us assume that this is 54%.
100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 0
Average of Males
sample size
10
20
30
40
50
60
70
80
90
100
Figure 1:
Evolution of average towards the mean value, as sample-size increases.
So, what is a parameter to a population is the corresponding statistic to a sample: The mean % value of males among the registered voters of Vienna is (known to be) 51.2% - a parameter of the population. The average % value of males among the 1,000 strong sample of city voters, is (measured to be) 54.0% - a statistic of the sample. Actually, and intuitively, the larger the sample size, the more reliable the conclusions are expected to be. So, as we increase the sample size, the average value is expected to converge to the mean value, i.e. the sample statistic becomes a more reliable estimate of the population parameter (see also Figure 1 above). The average value is a property of the specific sample, while the mean value is a property of the population. Needless to say that
Page 92 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
the average value is random: We get a different average every time we measure a new sample. Of course statisticians would be delighted to use the (entire) population, instead of a sample, to derive their conclusions, but such a desire is obviously not practical. Hence they have to settle for the next best thing which is to derive conclusions about the population, from a sample. So statisticians know how to estimate the reliability of their statistics, from the size and composition of their sample. They also know how to obtain useful estimates of the population parameters. Trend Analysis INHOPE hotlines collect monthly statistics of the reports they receive. When tabled, these statistics appear as below, where the 1st row gives the year and the month the statistic was collected and the 2nd row gives the statistic itself, for example the number of external reports received during the month.
2006- 2006- 2006- 2006- 2006- 2006- 2006- 2006- 2006- 200704 05 06 07 08 09 10 11 12 01
280
222
250
275
290
309
300
321
333
311
The 2nd row is a time-series, i.e. a set of numerical data collected regularly over time. The table above is normally represented in a simpler way:
1 2 3 4 5 6 7 8 9 10
280 222 250 275 290 309 300 321 333 311 The two tables are identical if one keeps a note that the 1st month corresponds to 2006-04. It is assumed that time series take on their values influenced by the following components22: 1. Trend: Overall or persistent, long-term upward or downward pattern of movement. It is maintained for several years and is due to
See for example “Basic Business Statistics”, by Berenson, Levine and Krehbiel, Pearson/Prentice-Hall, 2006.
Page 93 of 99 Not for publication without permission © INHOPE 2007
22
changes in public awareness, technology, population, wealth, values, etc. 2. Seasonal: Fairly regular periodic fluctuations that occur within a 12month period. They may be due to weather conditions, social customs, school schedules, etc. 3. Cyclical: Repeating up and down swings that are particularly longterm (2-10 years). 4. Irregular: Erratic fluctuations that are due to unsystematic reasons. The classical (multiplicative) time-series model for monthly data may thus be expressed as following:
where:
Yi = Ti x Si x Ci x Ii
Yi is the value of the time series (for example, the number of external reports) at month i (i=1,2,…), Ti is the value of the Trend component at month i, Si is the value of the Seasonal component at month i, Ci is the value of the Cyclical component at month i, Ii is the value of the Irregular component at month i, The aim of trend analysis is the determination of the trend component (Ti), or the filtering out of the other three components. Regarding the Trend component itself, there are a few models available, but in our case the following may only be considered: 1. Linear Trend Model: The parameter changes linearly with time, i.e., if for example the total number of reports increased by 200 over a period of 3 months, it does so over any period of three months, past, present, or future23. In a linear trend the speed of change of the parameter is constant. 2. Quadratic Trend Model: The rate at which the trend component changes, is not constant, like in the linear case, but it picks-up speed,
It must be stressed that, at this point in time, we do not propose that INHOPE would use trend analysis to forecast parameters of interest.
Page 94 of 99 © INHOPE 2007 Not for publication without permission
23
September 2007
Trend Report
or slows down. Frequently the trend exhibits a max (where it stops increasing and starts decreasing), or a min (where it does the opposite). In a quadratic trend the speed of change of the parameter is NOT constant, but the acceleration is constant. 3. Cubic Trend Model: The trend changes acceleration: For example it goes from an increasing speed to a decreasing speed of change (and vice versa). In a cubic trend neither the speed of change of the parameter nor the acceleration are constant. 4. 4th Order Polynomial Trend Model: Whilst the previous three models were in fact polynomial trend models of order 1, 2 & 3, respectively, the current one is a polynomial of order four. It contains a constant, a linear, a quadratic, a cubic and a 4th order component: Ti = A + Bxi + Cxi2 + Dxi3 + Exi4, [where i is the time unit and] A, B, C, D & E are constants to be determined so that we have the best fit (the error between the actual and the predicted is minimised). By comparison, the Cubic Trend Model is A + Bxi + Cxi2 + Dxi3. 5. Functions of polynomials: When a statistic is a function of other statistics for which we have the trend polynomials, the trend of the statistic under examination may be considered to be the (same) function of the contributing trend polynomials. Indices of quality We have described ways to obtain statistics, but we have not touched upon the main issue which is how we go from statistics to parameters, from samples to the population. To put it in a different way, when does one believe that the statistics obtained are an accurate enough estimate of the corresponding parameters? There must be a measure of reliability, an index of quality, which allows us to ‘believe’ the numbers and rely upon them, make decisions in other words. See for example the graph of Figure 2: A spreadsheet package readily gives trend-lines and their equations (Figure 3 & Figure 4). Do we ‘believe’ them though? Which one do we choose?
Page 95 of 99 Not for publication without permission © INHOPE 2007
80.000 70.000 60.000 50.000 40.000 30.000 20.000 10.000 0 0
Total Reports Received (EXT)
80.000 70.000 60.000 50.000 40.000 30.000 20.000 10.000 0
Total Reports Received (EXT)
y = 828,47x + 21630 R = 0,2087
0 5 10 15 20 25
2
5
10
15
20
25
Figure 2: Figure 3:
Monthly Time series of Total Reports Received (external) Monthly Time series of Total Reports Received (external), with linear trend-line and its equation
The parameter R2 displayed on the graphs is one index of quality. It tells us how close the time series is to the chosen model. 1-R2 is proportional to the average error squared24 between the actual time series and the model (so as this error tends to 0, 1-R2 tends to 0 and R2 tends to 1). In fact, the trend-line is chosen in such a way that R2 is maximized (or the average error is minimized).
80.000 70.000 60.000 50.000
400
Total Reports Received (EXT)
700 600 500
Y
y = 3,1627x + 143,67 R2 = 0,5652
40.000 30.000 20.000 10.000 0 0 5 10 15 20 25
300 200 100 0 0 20 40 60 80 100
y = 16,492x - 601,81x + 6844,5x + 7957 R = 0,3062
2
3
2
n
Figure 4: Monthly Time series of Total Reports Received (external), with cubic trend-line and its equation Figure 5: R2 (~57%) is low - Irregular component is strong. Linear fit is obvious though. There must be a measure of fit.
Figure 5 shows a time series with a low R2, but with an obvious linear trend. Hence there must be another index of quality. This is called significance and is denoted by “α”, which is the probability that the fit is wrong. Hence a low “α” is a good sign. In the case above α=0.000, hence the fit appears to be very good. In practice α is required to be <0.05, or <0.01. Points to Consider when deriving a trend line • We start with a scatter plot to observe the possible relationship between the parameter examined and time.
24 Note that the average error may be zero, without having a good fit, because negative and positive errors, however big in magnitude, may add up to zero – hence the need to add the squares of errors.
Page 96 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
•
We check the assumptions made by the theoretical models before these models are applied25: o Independence of errors: Error is the difference between the actual time series and the predicted (the trend model). When plotted against time the errors should not cluster but they must appear to be random. o Normality of errors: The errors must follow what is called a normal probability distribution. This is checked by plotting the errors on a special chart. o Equality of variance: Variance is a measure of how much a random variable differs from its mean value. It is required that the variance of errors is the same over all times, past and present. This is tested using a special method.
• •
If the assumptions are violated, alternative models should be used. If the assumptions are satisfied, tests are carried out about the significance of the results.
•
If
the
results
are
significant
they
are
quoted
together
with
appropriate confidence intervals. • While it is acceptable to interpolate the results and use the model to describe the ‘behaviour’ of the parameters within the time-period studied, it is dangerous to extrapolate, or attempt a forecast. If attempted, a forecast should not be done mechanically but must involve personal judgment, field experience and take into account changing technologies, habits, needs, etc. • We should be careful when attempting to explain the observed patterns of the time series: While causation implies correlation, correlation does not imply causation.
25
These assumptions are mentioned here for the first time.
Not for publication without permission © INHOPE 2007
Page 97 of 99
APPENDIX V LIST OF INHOPE MEMBERS (SEPTEMBER 2007)
(SEE WWW.INHOPE.ORG FOR CURRENT LIST)
Country Australia Austria Belgium Bulgaria Canada Chinese Taipei Cyprus Denmark Finland France Germany Germany Germany Greece Hungary Iceland Ireland Italy Italy Japan Malta Netherlands Poland Slovenia South Korea Spain United Kingdom United States
Organisation ACMA Stopline Child Focus ARC Fund cybertip.ca ECPAT Taiwan Safeweb Red Barnet STC Finland AFA ECO FSM jugendschutz.net SafeNet MATISZ Barnaheill ISPAI HOT 114 STC Italy Internet Association Japan APPOGG Meldpunt NASK Spletno Oko KISCOM Protegeles Internet Watch Foundation Cybertipline
Web Address www.au.inhope.org www.at.inhope.org www.be.inhope.org www.bg.inhope.org www.ca.inhope.org www.tw.inhope.org www.cy.inhope.org www.dk.inhope.org www.fi.inhope.org www.fr.inhope.org www.de.inhope.org www.de.inhope.org www.de.inhope.org www.gr.inhope.org www.hu.inhope.org www.is.inhope.org www.ie.inhope.org www.it.inhope.org www.it.inhope.org www.jp.inhope.org www.mt.inhope.org www.nl.inhope.org www.pl.inhope.org www.si.inhope.org www.kr.inhope.org www.es.inhope.org www.uk.inhope.org www.us.inhope.org
Page 98 of 99 © INHOPE 2007
Not for publication without permission
September 2007
Trend Report
Page 99 of 99
Not for publication without permission © INHOPE 2007