Docstoc

Mail-Filters_Technical_Presentation_v2

Document Sample
Mail-Filters_Technical_Presentation_v2 Powered By Docstoc
					Mail-Filters Technical
    Presentation

How it works, Why it’s Better
Mail-Filter Technology Overview
•   Why Mail-Filters
•   Bullet Signature Creation
•   Star Engine Process Overview
•   Implementation Options
•   SDK Contents
•   Getting Started
•   The API Commands
•   Testing Options
•   OEM Implementation Examples
•   FAQs
               Why Mail-Filters
• It’s Fast – 100s of messages per second (or higher)
• It’s Accurate – over 95% of spam caught, less than 1 in
  1,000,000 false positive rate
• Many implementation options – the right solution for any
  environment
• It’s Proprietary – it’s not fooled by spammer tricks - gives
  time to market and competitive differentiation
• It catches Foreign Language Spam – in over 30
  languages – a worldwide solution
• Easy Implementation – usually less than a day
• Full Support – Integration, technical support and training,
  marketing materials, sales training and lead generation
          How Mail-Filters Works
                        1. Spam Collection occurs
                           from many sources




                                5. Tuning Users and Administrators
                                provide feedback to help identify
                                spam and those that send them.

2. Human Editors Craft Bullet
Signatures




3. Bullet Signatures Are
Updated Every 1-15 Minutes        4. Mail-Filters Technology Integrated into
                                  OEM Solutions - Catches Spam, without False
                                  Positives
           Mail-Filters’ Process Overview To Capture
                Spam & Create Bullet Signatures
                                                                                                      Mail-Filters Data Centers

                                                    Bullet Signature Updater
               Customer
                                                                                                                                       Bullet
                              Customer submissions                                                                                   Signatures
                                                                          Traffic and Connection
  Mail-Filters                                                                   Heuristics
 Technology on
Customer Device                                                     Auto-Nominate
                                                                       Process

                                                                                                            Phish Trolling
                  International Spam                                    Spam          Culling
                                                                                                                                              Bullet
                       Harvester                                         DB           Engine
                                                                                                       Scam Sensors                         Signature
                                                                                                                                             Updates
                       www
                           www                                                                         Traffic Analysis
                  Phish Traps
                                                       Language
                                                       Assignment                                     Pre-Qualified &
                                                                               Prioritization
              Partner Collections                                                                     Auto-Nominated
                                                                                  Process
                                                                                                                                  Data Quality
                                                                                                         Reputation                Manager
                                                                                                          Analysis
                                                                              Aristotle
                       Spam Pre-Qualification                                (Signature
                                                                            Auto-Suggest)
  Partner Pre-Qualification
           Expert                                                                               Bullet Signature Creation
                                                                                                                                  Quality
                                                                                                Spammer Profile Creation
                                                                                                                                  Check
                                                Translation Tools
                                                                                                Message Profile Creation

                                                                Human Editors                        Traffic Profiles
      Star Engine Process Overview
                                                                                                           Mail-Filters
                                                                                                          Data Centers

                                                  STAR Engine Server
                                                                             Known Good Mail




                                           STAR Engine Management Module
                                                                                 Message
                                                                                Normalizer

                                                                             SnowFlake Buster
    OEM
                                                                                                   Bullet Signature
   Software                                                                  Language Analyzer         Updater
                                                                            Malformed Message
                        Is Message Spam?                                         Processor

Star Engine Interface                                                        Message Analysis
                          Yes / No

                                                                              Traffic Analysis           Bullet
                                                                                                       Signatures

                                                                            Reputation Analysis


                                                                           Spammer Profile Check

                                                                               False Positive
                                                                                Rationalizer
     Implementation Options
• Enterprise
  – Most typical implementation – highest
    performance – uses more resources
• Desktop
  – Small footprint – message is local – scan and
    database is remote
• Embedded
  – Tiny amount of resources required – scanning
    is done remotely
          Star Engine – Enterprise
                 (Very High Performance)
• Can process 100s or even over
                                       Server or Appliance
  1000 messages per second
                                           Hardware
• Requests Bullet Signature           Linked Together by OEM at compile
  updates every 1-10 minutes (only         OEM Application
  changes are downloaded)




                                                         C or C++
• The SEI and SES are typically




                                                                    API
  deployed on the same hardware
                                     Star Engine Interface (SEI)
• The SEI is linked into the OEM
  application using C or C++




                                                        TCP / IP
• The SES runs as a Service or
  Daemon and it manages it’s own     Star Engine Server (SES)
  Database Updates                          (Service or Daemon)

• The Database is usually between




                                                        TCP / IP
  3-10MB – will download a fresh
  DB upon startup if none present
                                               Mail-Filters
                                              Data Centers
      Star Engine - Enterprise
• The Star Engine Server is fully multi-threaded
• The Star Engine Server will run as a Service
  under Windows or as a Daemon under Linux,
  FreeBSD, or Solaris
• TCP/IP outbound on Port 80 is required – IP
  proxies are supported
• Typical requirements are P4, 100MB RAM, Hard
  Disk optional
• A unique Mail-Filters Customer ID is required to
  download the Bullet Signature Database
          Star Engine – Desktop
                  (Small Footprint)
• Only requires 128kb of RAM          PC or Other Device
• Can process 10s of messages          (with limited resources)
  per second                      Linked Together by OEM at compile

• Secondary server can be               OEM Application
  anywhere, including and




                                                     C or C++
                                                                API
  typically Mail-Filters’ Data
  Centers                             Star Engine Interface
• Database updates are not




                                                    TCP / IP
  required on the SEI (just the
  SES)                                  Separate Server
• Same exact API as the               Star Engine Server
  Enterprise implementation




                                                    TCP / IP
• Can also be used in a server
  cluster environment – many
  SEI’s feeding one SES                     Mail-Filters
                                           Data Centers
                 Star Engine – Embedded
                             A Completely New Approach
                                                                                      4. Mail-Filters’ authenticates as the
                                                                                      user to the ISP or Corporate email
•   Anti-Spam detection for edge devices with      Email Server                       servers - the mail is delivered
    almost no resource requirements
•   OEM code requires less than 10kb of
    RAM
•   No software need be installed on any user
    PC – the service is turned on or off at the               3, Mail-Filters makes the request on
                                                              behalf of the user, filters the
    OEM device                                                messages, then sends the good mail
•   Works with POP3 & IMAP                                    to the user. No mail is kept at Mail-
                                                              Filters – it just passes through.
•   OEM device intercepts the message
    delivery request and sends it to Mail-
    Filters                                                                         WWW                           Mail-Filters
                                                                                                                  Data Centers
•   Mail-Filters receives the messages on
    behalf of the end user, filters for viruses
    and spam, then sends the clean                2. OEM device intercepts the request
    messages to the end user                      based on port the request is made on
•   OEM or customer determines what               (Ex. 110 = POP3) – and redirects the
                                                  request to Mail-Filters’ data centers.
    happens to spam (delete, mark with an X-
    header, decorate the subject line)
•   Since spam can be deleted and the
    downlink speed is probably slower than
    the link from Mail-Filters’ data centers to
    the email servers – good mail will get to
    the end user faster.
                                                                                       1. Email Client requests
                                                                                                                       PC
                                                                                       mail
                 Embedded Architecture
                                                                                Email Server
                                                                                                     Mail-Filters
                                                                                                     Data Centers


                                                                                The Internet


                                                OEM Device



                                                                                      Redirect Code
                              OEM Application


     Outbound Listening Code (Port 110 for POP3 or Port 147 for IMAP Requests)




               Customer                 The Email Client requests email from an email server – it makes the request on
               Premise                  port 110 or 147 – the OEM device redirects the request to Mail-Filters. A port is
                                        opened by the email server via Mail-Filters to the PC. The email is filtered, a
                             PCs        policy is applied, then delivered to the Email Client.
Email Server
             SDK Contents
• Star Engine Server software executables
• Star Engine Interface libraries in C and
  C++
• Simple Single-Threaded implementation
  example application
• Documentation

• Typical integration time is less than a day
  Getting Started with the SDK
• Install the Star Engine Server
• Run the Star Engine Server
• Run the Example Application
  – This application will scan the files in the
    directory of choice and all sub-directories to
    see if they are spam. The results will display
    on the screen.
• Begin the Integration to the OEM
  application
           The Star Engine API
            (The Star Engine Interface)
• The Commands are Straight-Forward
  – Initialize – This command establishes a connection to
    the Star Engine Server
  – Shutdown – Used to tear down the thread after a
    successful Initialize command
  – Scan SMTP Buffer – Passes the SES the data to be
    scanned – will return TRUE if Spam
  – SCAN Buffer – Passes the SES data to be scanned –
    best used for non-SMTP types of content such as IM,
    SMS, web pages, etc.
  – Version – Returns the versions of all the components
    currently being used, including the database version
    date.
                     Testing Options
• The Mail-Filters database is culled to eliminate old/unused signatures.
    – As a result, the catch rate will suffer on old corpuses of email
    – Best results are obtained with live (or very close to it) email.
• There are several options to test the Mail-Filters technology
    – To test for catch rate or false positive rate
        • Use the Example scan utility to check individual messages in a directory
        • Send mail to an account Mail-Filters can set up for you at Cleantree.com.
          Good mail will go to the Inbox, spam to the Spam folder. Check results using
          your browser.
        • Integrate into the OEM application and run it to check catch rate.
    – To test throughput:
        • Unfortunately, the Example application is only a single-threaded application
          and will not show what the SES can achieve throughput-wise (it does fine on
          catch rate)
        • The only fair test is to do an integration and run email through it. Most OEMs
          fine the solution throughput is the same whether Mail-Filters technology is
          running or not.
    – To test Foreign Language:
        • Do a beta test with a customer or partner in the region of interest
        • Mail-Filters have several partners in various regions that may assist in a beta
          test, if desired.
     Implementation Examples
• Enterprise
  – Most OEMs have implemented the Mail-Filters
    technology as the primary anti-spam solution
     • AV solutions company scans for spam while it has the message
       in memory to scan for viruses. Because spam is more prevalent
       and is a much faster scan, spam is typically scanned for first.
  – Some have augmented their own anti-spam technology
     • Because Mail-Filters technology is both fast and accurate, some
       have used it as a pre-processor to their own, more
       computationally expensive technology, to increase the
       throughput of the overall solution, and to increase spam catch
       rates.
    Implementation Examples
• Desktop
  – Some devices don’t have the processing power or
    resources available for spam detection. For these,
    the Mail-Filters technology can provide a smaller
    footprint
     • Firewalls, security gateways, messaging gateways,
       enterprise PCs may prefer a secondary server to handle the
       scanning to free up resources on their own hardware.
  – An MSP has a cluster environment where there are
    many SEIs feeding one SES per tower. This is very
    efficient and allows their overall throughput to
    increase dramatically.
     Implementation Examples
• Embedded
 – Ideal for DSL routers, Cable Modems, Wireless
   gateways, SMB security gateways etc.
 – Because it requires no end user software
   installation or configuration, it is simple to sign-
   up and have spam and viruses eliminated.
   Frequently Asked Questions
• How do I get the SDK?
   – Sign the Mail-Filters MNDA and we’ll send it to you via email.
• Is the Star Engine Server multi-threaded?
   – Yes.
• Does it handle messages in double-byte character sets?
   – Yes, our technology catches spam in over 30 languages,
     including multi-byte character sets such as Japanese, Korean,
     Chinese, Arabic, and Hebrew.
• How is the update interval set – can it be changed?
   – The update interval is set by the OEM, but can be changed on a
     customer by customer basis. The default is an incremental
     every 10 minutes and a full update written to disk once a week.
• Will this solution work on less than a Pentium IV PC?
   – Yes, but it works more efficiently on a PIV.
     Frequently Asked Questions
•   What happens if the SES can’t get a database, or quits running, or some
    other catastrophe?
     – The SES or SEI will fail safe. It will return a FALSE ( the message isn’t spam)
       and continue to process messages while trying to reconnect. The customer will
       see more missed spam, but won’t miss any messages.
•   What if the SES doesn’t have the rights to write the database to disk, or the
    disk is full?
     –   The SES will continue to function properly and will acquire updates to the
         database in memory. The version command will return the database currently
         being used in RAM.
•   Is the API really just 5 functions?
     – Yes – it doesn’t get much simpler than that.
•   Can the SES return a probability of a message being spam?
     – No - Because the technology uses human editors to craft profiles and message
       signatures, we’re very very confident the message is spam if we identify it.
       Because our false positive rate is so low, our methodology is proven to be
       correct. A probability is required by technologies that guess or compute whether
       a message is spam – we know it, so we tell you. For those solutions that require
       a probability, they set our TRUE response to the highest probability – 10 or 1 or
       100.
               Conclusions
• The Mail-Filters technology is easy to
  implement and provides options for any
  situation.
• The underlying technology far surpasses
  what others are doing, giving the Mail-Filters
  OEM a significant advantage over
  competitors in catch rate and accuracy,
  language coverage, and throughput.
• Human review provides the difference -the
  technology delivers it.

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:5
posted:10/27/2011
language:English
pages:22