chapter11 by ashrafp


									   Agent Technology for e-Commerce

        Chapter 11: Mobile Agents

                       Maria Fasli
Agent Technology for e-Commerce

                              Current approaches
   Current approaches to distributed computing:
    The client-server model

    Code-on-demand

    Web services approach

      Advantages
         security

      Disadvantages
         network traffic

         latency

         overloading of resources

Chapter 11                                         2
Agent Technology for e-Commerce

             An alternative approach: mobile agents
      An autonomous executing entity that has the ability to migrate
       from machine to machine in a heterogeneous network and resume
       its execution. They can:
      reduce network traffic
      overcome network latency
      make better use of resources
      execute asynchronously and autonomously
      adapt dynamically
      be robust and fault tolerant

Chapter 11                                                          3
Agent Technology for e-Commerce

                       Infrastructure requirements
   To implement a mobile agent we need:
    a suitable and transportable language in which the agent’s code
     will be written,
    an engine or interpreter for that language,

    communication protocols that allow engines on different
     machines to exchange agents.



Chapter 11                                                             4
Agent Technology for e-Commerce


                Sender host                       Receiver host
                    Engine                              Engine
                       Suspension                   Execution resumption

                       Serialization                     Deserialization

                        Encoding                           Decoding

                         Dispatch                           Receipt

                                       Agent transfer

Chapter 11                                                                 5
Agent Technology for e-Commerce

                              Modes of migration
   The statefull or strong mobility (go) model
    An agent’s object state, code, and control state are captured.
     Execution is allowed to continue from the exact point at which it
     was stopped.
    More convenient for the end programmer, but more work for the
     system developer. Routines to capture state control are required
     on top of the interpreters.
   The stateless or weak mobility (known entry point) model
    The agent’s object state and code is captured. The execution in
     the new machine is continued from a known entry point.
    Commercial Java-based systems use the entry point.

Chapter 11                                                               6
Agent Technology for e-Commerce

                            Mobile agent systems
   Non-Java mobile agent systems
       Telescript

       Agent Tcl

       D’Agents

       Ara

       TACOMA

   Java-based mobile agent systems
       Aglets

       Concordia

       NOMADS

Chapter 11                                         7
Agent Technology for e-Commerce

      Implemented at the IBM Tokyo Research Laboratory.
      Aglets do not capture an agent’s control state during migration.
       That would require modifications to the standard Java virtual
       machine. Instead agent execution is restarted from a known entry

Chapter 11                                                            8
Agent Technology for e-Commerce

                                  Basic concepts
      Aglet: a mobile Java object that visits aglet-enabled hosts in a
      Proxy: it is a representative of an aglet. It serves as a shield to
       protect the aglet from direct access to its public methods.
      Context: an aglet’s workplace. It corresponds to a place. Multiple
       contexts may reside on the same machine.
      Identifier: a unique identifier is assigned to each aglet after

Chapter 11                                                               9
Agent Technology for e-Commerce

                    Fundamental Aglet operations
      Creation
      Cloning
      Dispatching
      Retraction
      Activation and Deactivation
      Disposal

Chapter 11                                         10
Agent Technology for e-Commerce

                             Programming model
      The Aglet programming model is event-based
      Customised listeners can be used to catch particular events in the
       life cycle of an aglet:
         Clone listener (clone creation)

         Mobility listener (dispatching, retraction, arrival)

         Persistence listener (deactivation, activation)

Chapter 11                                                              11
Agent Technology for e-Commerce

   General process
    A mobile agent that wants to migrate calls the dispatch
    The Aglets system calls the agent’s onDispatching method
     which performs application specific cleanup, kills the agent’s
     threads, serialises the agent’s code and object state, and sends the
     agent’s code and object state to the new machine
    On the new machine the system calls the agents onArrival
     method which performs application specific initialisation
    Finally the agent’s run method is called in order to start the
     agent’s execution

Chapter 11                                                             12
Agent Technology for e-Commerce

      Communication between aglets
         When an aglet wants to communicate with other aglets, it first

          has to obtain a proxy
         Public methods cannot be accessed directly

         Three different types of messages: now-type, future-type,


      Communication between engines
         The ATP is used in the communication layer

         The API abstracts the communication between agent systems

         Methods for creating, transferring, tracking and managing


Chapter 11                                                            13
Agent Technology for e-Commerce

                                  Security model
      The first level of security comes from Java itself. Imported code
       fragments are subjected into a series of tests and consistency tests
      As the second level the Tahiti (the visual agent manager)
       implements a configurable security manager that provides a fairly
       high degree of security for the hosting computer and its owner
       and it enables a number of principals to specify and enforce

Chapter 11                                                               14
Agent Technology for e-Commerce

                            Mobile agent security
   Mobile agent security has two aspects:
    Protecting the host machine

    Protecting the mobile agent

   Mobile agents and hosts are vulnerable to a number of threats
    Agent to host threats

    Agent to agent threats

    Host to agent threats

Chapter 11                                                         15
Agent Technology for e-Commerce

                             Agent to host threats
      Masquerade
      Unauthorized access
      Tampering
      Denial of service

Chapter 11                                           16
Agent Technology for e-Commerce

                                  Agent to agent
      Masquerade
      Tampering
      Denial of service
      Repudiation

Chapter 11                                         17
Agent Technology for e-Commerce

                                  Host to agent
      Masquerade
      Tampering
      Eavesdropping and traffic analysis
      Denial of service

Chapter 11                                        18
Agent Technology for e-Commerce

                                  Security services
      Authentication
         User

         Host

         Agent

         Code

      Access control
      Integrity
      Confidentiality
      Nonrepudiation
      Auditing

Chapter 11                                            19
Agent Technology for e-Commerce

                                  Protecting the host
   A number of approaches have been suggested:
    Safe code interpretation: use of an interpreted language

    Authentication: digital signatures

    Authorization: entities can be granted capabilities or access rights

    Resource allocation: various mechanisms for resource allocation,
     including market mechanisms
    Maintaining path histories in order to be able to check the origin
     of mobile agents

Chapter 11                                                             20
Agent Technology for e-Commerce

                       Protecting the mobile agent
      Fortress model: agents execute in a safe environment where no
       untrusted hosts are allowed – compromises the advantages of
       mobile agents
      Sealing and signing intermediate results to prevent tampering
      Using shared secrets and interlocking
      The agent’s execution and behaviour is recorded (Execution
      Executable encrypted functions can prevent tampering

Chapter 11                                                             21
Agent Technology for e-Commerce

                          Issues on mobile agents
   Technical issues
    Current mobile agents do not bring about significant benefits

    As they are written in interpreted languages, they are slow

    Starting an execution environment and inserting an agent
     involves an overhead – thus higher loads in transporting and
     executing them locally
    To recover from failure during migration additional support is
    Appropriate naming and location services are required

    Security

    Lack of standardization

Chapter 11                                                            22
Agent Technology for e-Commerce

   Nontechnical issues
    Lack of a killer application

    The advantages of mobile agents are modest when applications
     are considered in isolation

   Potential applications
    Electronic marketplaces

    Mobile devices such as mobile phones and PDAs

Chapter 11                                                          23

To top