Multi Agent Based De-Centralized
               Knowledge Discovery and Agent
                      Security: A Review

                                      Presentation by:
                                      Aman Kumar
                                      M.tech -CSE (2nd Sem)
                                     Graphic Era University,
                                      Dehradun,India
May 26, 2012                                                   1
Agenda
    Introduction
    Data Mining Vs. De-Centralized Data Mining
    Agent, Why Agent?
    Agent Based De-Centralized Data Mining
    MADM Systems-An Architectural Approach
    Advantages of MADM
    Agents Security Issues
    Security measures for agent
    Future Scope
    Summary
    References

May 26, 2012                                      2
Introduction

    Data Mining-KDD
    De-Centralized Environment
    De-Centralized KDD(DDM)
    An Agent
    Multi Agent System




May 26, 2012                      3
Data Mining
      General   Data Mining Model


                       Data Mining   Data
                       Tools         Warehouse




May 26, 2012                                     4
Distributed Data Mining
       De-Centralized        Data Mining model


                         Local Model Aggregation                 Final Model



               Local                                Local         Local
               Model                                Model         Model




               Data Mining                         Data Mining   Data Mining
               Algorithm                           Algorithm     Algorithm




May 26, 2012                                                                   5
                Site k                               Site 2        Site 1
Agent , Why Agent?
   An Software Agent is as user’s personal
    assistant .
   Agent can be programmed as compact as
    possible.
   Light weight agent can transmitted across the
    network rather than data that is more bulky.
   The Designing of DDM Systems Deals With
        Great Details of Algorithms used
        Reusability
        Extensibility
        Robustness

   Hence , the agent characteristics are desirable
    to use in DDM.
May 26, 2012                                          6
Agent Based DDM
              ADDM system concerns three keys characteristics

                    Interoperability
                    Dynamic Configuration
                    Performance Aspects

              Application’s of distributed data mining include credit card
               Authentication, intrusion detection and all this type general
               and security related applications.
              Into this a novel Data Mining Technique inherits the
               properties of agents.
              The DDM applications can be further enhanced with agents.
              Better Integration policy with the communication protocols
              Provide a view of online parallel processing


May 26, 2012                                                                   7
Basic Components of ADDM
              An ADDM system can be generalized into a set of
               components




                                                   Application
                                                   Layer


                                                  Data Mining
                                                  Layer



                                                  Agent Grid
                                                  Infrastructure
                                                  Layer

                   Fig. Overview of ADDM
May 26, 2012                                                       8
MADM Systems-
 An Architectural Approach
         MADM is the ADDM but equipped with several agents which
          have particular goal of functionality as:

         Resource Agent: Maintaining Meta Data Information
         Local Task Agent: Located at the local site
         Broker Agent: Working as Advisor agent
         Query Agent: KDD System Agent
         Pre-Processing Agent: Preparing data for mining
         Post Data Agent: Evaluates the performance and accuracy
         Result Agent: Aggregate the all local results
         Interface Agent: Provide Interface to the real world
          applications
         Mobile Agent: Migrate based on Request and Response

May 26, 2012                                                        9
MADM Systems-An Architectural
Approach contd….




                       Data Source
                       On Different
May 26, 2012
                       Sites          10
Agent Security Issues
    Identification and authentication
    Authorization and delegation
    Communication
     • confidentiality: assurance that communicated information is not
        accessible to unauthorised parties;
     • data integrity: assurance that communicated information cannot be
     manipulated by unauthorised parties without being detected;
     • availability: assurance that communication reaches its intended
     recipient in a timely fashion;
     • non-repudiation: assurance that the originating entity can be held
     responsible for its communications.
    Mobility
    Situated ness
    Autonomy
    Agent Execution

May 26, 2012                                                                11
Security Measures for agent
    Protecting agents
    Trusted hardware
    Trusted nodes
    Co-operating agents
    Execution tracing
    Encrypted payload
    Environmental key generation
    Computing with encrypted functions
    Un-detachable signatures




May 26, 2012                              12
Security Measures ……
    Protecting the agent platform

    Sandboxing and safe code interpretation
    Proof carrying code
    Signed code
    Path histories

     State appraisal




May 26, 2012                                   13
Future Scope
        Data Mining and web mining is the hot area of research
        Integration of KDD and Agent technology can provide a new
         way to both
        For several network security researchers it can provide
         several new way to find the fraud in the network as provide
         fast discovery
         Real time confidential transaction can be make secure by
         the integration of Agent Technology




May 26, 2012                                                           14
Summary

         This presentation presented an overview of :-
            Data Mining

            Distributed Data Mining

            Agent Based DDM

            MADM systems as survey based on the information that
             are exist today.
            common components between these systems and gives a
             description to their strategies and architecture.
            Security Measures For Agents

         This presentation shows the integrated architectural model
         of distributed data mining and the agent technology, which
         provide a optimized performance to the knowledge
         discovery when the data is not resides in a central site or
         scattered over the network.
May 26, 2012                                                           15
References
          [1] IJRIC ISSN: 2076-3328 www.ijric.org E-ISSN: 2076-3336 “. Agent based distributed data
          mining: AN OVER VIEW “VUDA SREENIVASA RAO, 2009-2010
         [2] M. Klusch, S. Lodi, G. Moro. Agent-based Distributed Data Mining: The KDEC Scheme.
          Intelligent Information Agents - The AgentLink Perspective. Lecture Notes in Computer Science 2586
          Springer 2003.
         [3] “Distributed Data Mining and Agents” Josenildo C. da Silva, Chris Giannella, Ruchita Bhargava,
          Hillol Kargupta1;, and Matthias Klusch
         [4] Y. Xing, M.G. Madden, J. Duggan, G. Lyons. A Multi-Agent System for Context-based Distributed
          Data Mining. Technical Report Number NUIG-IT-170503, Department of Information Technology,
          NUI, Galway, 2003.
         [5] “Agent-Based Data-Mining” Winton Davies 15 August 1994
         [6] Priyanka Makkar et. al. / (IJCSE) International Journal on Computer Science and Engineering
          Vol. 02, No. 04, 2010, 1237-1244 DISTRIBUTED DATA MINING AND MINING MULTI-AGENT
          DATA ,Vuda Sreenivasa Rao, Dr. S Vidyavathi
         [7] V. Gorodetsky and I. Kotenko. “The Multiagent Systems for Computer Network Security
          Assurance: frameworks and case studies.” In IEEE International Conference on Artificial Intelligence
          Systems, 2002, pages 297–302, 2002.
         [8] International Journal of Computer Applications (0975 – 8887) Volume 4– No.12, August 2010
         23 “A Comparative study of Multi Agent Based and High- Performance Privacy Preserving Data
          Mining”, Md Faizan Farooqui, Md Muqeem, Dr. Md Rizwan Beg
         [9] Future Generation Computer Systems 23 (2007) 61–68 ,www.elsevier.com/locate/fgcs
          “Distributed data mining on Agent Grid: Issues, platform and development toolkit” Jiewen Luoa,b,_,
          Maoguang Wangc, Jun Hud, Zhongzhi Shia
         [10] Sung W. Baik, Jerzy W. Bala, and Ju S. Cho. Agent based distributed data mining. Lecture
          Notes in Computer Science, 3320:42–45, 2004.
         [11]Xining Li and Jingbo Ni. Deploying mobile agents in distributed data mining. Lecture Notes in
          Computer Science 4819:322–331, 2007 .
May 26, 2012                                                                                                     16
    [12]“Mobile agent security” Niklas Borselius Mobile VCE Research Group Information Security Group,
    Royal Holloway, University of London Egham, Surrey, TW20 0EX, UK ,Niklas.Borselius@rhul.ac.uk
Thank You



May 26, 2012               17

Distributed Datamining and Agent System,security

  • 1.
    Multi Agent BasedDe-Centralized Knowledge Discovery and Agent Security: A Review Presentation by: Aman Kumar M.tech -CSE (2nd Sem) Graphic Era University, Dehradun,India May 26, 2012 1
  • 2.
    Agenda  Introduction  Data Mining Vs. De-Centralized Data Mining  Agent, Why Agent?  Agent Based De-Centralized Data Mining  MADM Systems-An Architectural Approach  Advantages of MADM  Agents Security Issues  Security measures for agent  Future Scope  Summary  References May 26, 2012 2
  • 3.
    Introduction  Data Mining-KDD  De-Centralized Environment  De-Centralized KDD(DDM)  An Agent  Multi Agent System May 26, 2012 3
  • 4.
    Data Mining General Data Mining Model Data Mining Data Tools Warehouse May 26, 2012 4
  • 5.
    Distributed Data Mining De-Centralized Data Mining model Local Model Aggregation Final Model Local Local Local Model Model Model Data Mining Data Mining Data Mining Algorithm Algorithm Algorithm May 26, 2012 5 Site k Site 2 Site 1
  • 6.
    Agent , WhyAgent?  An Software Agent is as user’s personal assistant .  Agent can be programmed as compact as possible.  Light weight agent can transmitted across the network rather than data that is more bulky.  The Designing of DDM Systems Deals With  Great Details of Algorithms used  Reusability  Extensibility  Robustness  Hence , the agent characteristics are desirable to use in DDM. May 26, 2012 6
  • 7.
    Agent Based DDM  ADDM system concerns three keys characteristics  Interoperability  Dynamic Configuration  Performance Aspects  Application’s of distributed data mining include credit card Authentication, intrusion detection and all this type general and security related applications.  Into this a novel Data Mining Technique inherits the properties of agents.  The DDM applications can be further enhanced with agents.  Better Integration policy with the communication protocols  Provide a view of online parallel processing May 26, 2012 7
  • 8.
    Basic Components ofADDM  An ADDM system can be generalized into a set of components Application Layer Data Mining Layer Agent Grid Infrastructure Layer Fig. Overview of ADDM May 26, 2012 8
  • 9.
    MADM Systems- AnArchitectural Approach  MADM is the ADDM but equipped with several agents which have particular goal of functionality as:  Resource Agent: Maintaining Meta Data Information  Local Task Agent: Located at the local site  Broker Agent: Working as Advisor agent  Query Agent: KDD System Agent  Pre-Processing Agent: Preparing data for mining  Post Data Agent: Evaluates the performance and accuracy  Result Agent: Aggregate the all local results  Interface Agent: Provide Interface to the real world applications  Mobile Agent: Migrate based on Request and Response May 26, 2012 9
  • 10.
    MADM Systems-An Architectural Approachcontd…. Data Source On Different May 26, 2012 Sites 10
  • 11.
    Agent Security Issues  Identification and authentication  Authorization and delegation  Communication • confidentiality: assurance that communicated information is not accessible to unauthorised parties; • data integrity: assurance that communicated information cannot be manipulated by unauthorised parties without being detected; • availability: assurance that communication reaches its intended recipient in a timely fashion; • non-repudiation: assurance that the originating entity can be held responsible for its communications.  Mobility  Situated ness  Autonomy  Agent Execution May 26, 2012 11
  • 12.
    Security Measures foragent  Protecting agents  Trusted hardware  Trusted nodes  Co-operating agents  Execution tracing  Encrypted payload  Environmental key generation  Computing with encrypted functions  Un-detachable signatures May 26, 2012 12
  • 13.
    Security Measures ……  Protecting the agent platform  Sandboxing and safe code interpretation  Proof carrying code  Signed code  Path histories  State appraisal May 26, 2012 13
  • 14.
    Future Scope  Data Mining and web mining is the hot area of research  Integration of KDD and Agent technology can provide a new way to both  For several network security researchers it can provide several new way to find the fraud in the network as provide fast discovery  Real time confidential transaction can be make secure by the integration of Agent Technology May 26, 2012 14
  • 15.
    Summary  This presentation presented an overview of :-  Data Mining  Distributed Data Mining  Agent Based DDM  MADM systems as survey based on the information that are exist today.  common components between these systems and gives a description to their strategies and architecture.  Security Measures For Agents  This presentation shows the integrated architectural model of distributed data mining and the agent technology, which provide a optimized performance to the knowledge discovery when the data is not resides in a central site or scattered over the network. May 26, 2012 15
  • 16.
    References  [1] IJRIC ISSN: 2076-3328 www.ijric.org E-ISSN: 2076-3336 “. Agent based distributed data mining: AN OVER VIEW “VUDA SREENIVASA RAO, 2009-2010  [2] M. Klusch, S. Lodi, G. Moro. Agent-based Distributed Data Mining: The KDEC Scheme. Intelligent Information Agents - The AgentLink Perspective. Lecture Notes in Computer Science 2586 Springer 2003.  [3] “Distributed Data Mining and Agents” Josenildo C. da Silva, Chris Giannella, Ruchita Bhargava, Hillol Kargupta1;, and Matthias Klusch  [4] Y. Xing, M.G. Madden, J. Duggan, G. Lyons. A Multi-Agent System for Context-based Distributed Data Mining. Technical Report Number NUIG-IT-170503, Department of Information Technology, NUI, Galway, 2003.  [5] “Agent-Based Data-Mining” Winton Davies 15 August 1994  [6] Priyanka Makkar et. al. / (IJCSE) International Journal on Computer Science and Engineering Vol. 02, No. 04, 2010, 1237-1244 DISTRIBUTED DATA MINING AND MINING MULTI-AGENT DATA ,Vuda Sreenivasa Rao, Dr. S Vidyavathi  [7] V. Gorodetsky and I. Kotenko. “The Multiagent Systems for Computer Network Security Assurance: frameworks and case studies.” In IEEE International Conference on Artificial Intelligence Systems, 2002, pages 297–302, 2002.  [8] International Journal of Computer Applications (0975 – 8887) Volume 4– No.12, August 2010  23 “A Comparative study of Multi Agent Based and High- Performance Privacy Preserving Data Mining”, Md Faizan Farooqui, Md Muqeem, Dr. Md Rizwan Beg  [9] Future Generation Computer Systems 23 (2007) 61–68 ,www.elsevier.com/locate/fgcs “Distributed data mining on Agent Grid: Issues, platform and development toolkit” Jiewen Luoa,b,_, Maoguang Wangc, Jun Hud, Zhongzhi Shia  [10] Sung W. Baik, Jerzy W. Bala, and Ju S. Cho. Agent based distributed data mining. Lecture Notes in Computer Science, 3320:42–45, 2004.  [11]Xining Li and Jingbo Ni. Deploying mobile agents in distributed data mining. Lecture Notes in Computer Science 4819:322–331, 2007 . May 26, 2012 16 [12]“Mobile agent security” Niklas Borselius Mobile VCE Research Group Information Security Group, Royal Holloway, University of London Egham, Surrey, TW20 0EX, UK ,Niklas.Borselius@rhul.ac.uk
  • 17.