1. AT2AI-4
Fourth International Symposium "From Agent Theory to Agent Implementation"
An Ontological Approach to Harmonising Security Models for Open Services
Juan Jim Tan (Presenter), Stefan Poslad, Leonid Titkov
{ juanjim.tan, stefan.poslad, leonid.titkov

2. Presentation Outline Background & Motivation
Related Work and its Limitations
Holistic Security Ontology Model
Conceptual & Service Description Layers
Policy & Application Layers
Critical Analysis
Evaluation of Conceptual Framework
Evaluation of Reasoning Model Performance
Conclusion and Future Work

3. Background Security concerns of open societies
More vulnerabilities
Insider attacks
Balancing the benefits of open trading vs. closed system security
Increasing ubiquitous infrastructure, e.g., Semantic Web:
Many Open Heterogeneous Services
Services Communicating and Interoperating
Service Composition
Virtual Enterprise / Virtual Organisation
Security models for open societies
Different application models, community self-policing
These complement conventional e-Business security for Payment, protection of systems, and etc.

4. Motivation Plethora of security standards consortia
Difficult to agree on a single standard
Different standards have its limitations
Security capabilities unstructured
Different islands of security implementations
Lack of explicit method
No common terminology of stakeholders
Abstraction permits propagation of low level faults to other parts of the system
Difficult to propagate low-level faults in one part of the system to another part at a high enough level of abstraction to make decisions about the cause and effect of multiple heterogeneous low-level faults

5. Objectives
“Describe how distributed security issues can be approached in Open Multi Agent Multi-domain (MAMD) Environments in which applications can dynamically interoperate securely “
“A holistic framework capturing stakeholders and security information within an open service environments using Dynamic Reconfiguration without affecting underlying implementations”

6. Research Contributions
Holistic Abstract and Explicit Framework for Securing Open Services:
V-SAT Conceptual Model – Upper Ontology
Decentralised advertising and discovering of security profiles
Dynamic reconfiguration – Policy based approaches
Risk Management Model
Automated collaboration and reasoning
Operational Model

8. Presentation Outline Background & Motivation
Related Work and its Limitations
Holistic Security Ontology Model
Conceptual & Service Description Layers
Policy & Application Layers
Critical Analysis
Evaluation of Conceptual Framework
Evaluation of Reasoning Model Performance
Conclusion and Future Work

9. Comparison of Security Specifications
IETF, Oasis, and W3C

10. Comparison between V-SAT, KAoS, Rei, Ponder
Ontology- based
Yes No
Policy Type
Constraint or Pre-condition based
Access Control based
Policy
Representation
KIF + RDF(-S)
*developing support for other representations
OWL
Rei: ( Prolog-like syntax + RDF-S)
Ponder language specification
Open Security Interoperability Support
Security Profiling – capable of representing security instances and complex processes of open semantic based systems
Applying mediating and proxy agents onto specific domains
Applicable in specific domains
Interoperability Representation
Abstract to Explicit
Explicit
Reasoning support
Java Theorem Prover
Prolog engine
Event calculus representation

11. Comparison - V-SAT, WS-Security, GRID Security
Ontology- based
Yes – Semantic
No – Syntactic No
Policy Support
Yes – Constraint
Yes – Authorizations
Policy Representation
KIF + RDF(-S)
XML
Grid policy / WS-Security
Interoperability Support
Yes – Security Profiles
Security Token - Claims
Applicable in GRID environment / Utilise WS-Security Spec.
Interoperability Representation
Abstract (Model) to Explicit
Specification
Explicit
Reasoning support
Java Theorem Prover
Not defined
Risk Management
Yes
Analysability
Ontology simplifies reasoning and dependency validation
Specific to a particular context/domain and lacking in inter-concept validation
Expressivity
Represent behavior of complex environment, multi level abstractions, and easy to extend new concepts
Represent only specific behaviors and difficult to extend new concepts
V-SAT also supports ‘spatial’ and ‘temporal’ event conditions within the reasoning model.

12. Other Related Work [Denker et al., 2003];
Focuses on DAML based implementation
Limited conceptual model
Complicated negotiations – brittle in large scale systems

13. Presentation Outline Background & Motivation
Related Work and its Limitations
Holistic Security Ontology Model
Conceptual & Service Description Layers
Policy & Application Layers
Critical Analysis
Evaluation of Conceptual Framework
Evaluation of Reasoning Model Performance
Conclusion and Future Work

14. Holistic Abstract Security Model
Security Applications
Policies
Descriptions
Service
Trust
Security Ontology Model
Concepts
Security
Trust
Privacy
Security Mechanisms and standard specifications, e.g., SSL, XML-Signature etc.

15. Profile Driven Open Service Security Framework

16. Service Composition Scenario

17. Conceptual Layer – Semantic Model

18. Service Description Layer 1/2
Advertised into Directory Services for discovery, e.g. LDAP, UDDI, DF, etc

19. Service Description Layer – Risk Model 2/2

20. Presentation Outline Introduction & Motivation
Related Work and its Limitations
Holistic Security Ontology Model
Conceptual & Service Description Layers
Policy & Application Layers
Critical Analysis
Evaluation of Conceptual Framework
Evaluation of Reasoning Model Performance
Conclusion and Future Work

21. Policy Layer – Reasoning Model 1/2

22. Policy Layer – Reasoning Model 2/2
Reasoning Algorithm Steps using JTP:
Loading of Ontologies, General Rules, User and Service Profiles
Spatial and Temporal Threat Detection; checks if environment is trusted (end) or un-trusted:
Identify associated threats and safeguards
Compare user and service profile instances and policies
If there are conflicts – resolve conflicts by order and precedence of policies
Result presenting the outcome of the reasoning process for security decision making
Identify associated threats and safeguards to determine a rational recommendation or decision making over what security instances and policies to support.

23. Application Layer – Operational Model
Modular security API developed

24. Presentation Outline Introduction & Motivation
Related Work and its Limitations
Holistic Security Ontology Model
Conceptual & Service Description Layers
Policy & Application Layers
Critical Analysis
Evaluation of Conceptual Framework
Evaluation of Reasoning Model Performance
Conclusion and Future Work

25. Why our modelling choice?
Reuse current security models as an upper ontology for supporting interoperability
We can rely on current best security practises
Able to capture stakeholders in an abstract way that can be strapped onto an explicit model
Interoperability is more achievable using semantic approaches to practical applications

26. Why an Explicit Security Model?
To define explicit security requirements, protection and policies of systems as profiles
Argument:
Security should not be advertised because it gives an inside information of the system;
Security by obscurity gives a false sense of security
Hide weaknesses that an open peer review might uncover

27. Framework Achievements
Deployed Security Applications in a Global Agent Test-bed (Agentcities.RTD) project.
Ontology and specifications available at
Reasoning Model implemented using JTP, DAML+OIL and KIF
Developed a Risk Model that assesses the criticality and probability of threats
Web based demo of an Open Security Interoperability Scenario (to be released)

28. Presentation Outline Introduction & Motivation
Related Work and its Limitations
Holistic Security Ontology Model
Conceptual & Service Description Layers
Policy & Application Layers
Critical Analysis
Evaluation of Conceptual Framework
Evaluation of Reasoning Model Performance
Conclusion and Future Work

29. Performance of Loading Facts and Rules into JTP
Checkpoints: 1 = Security Ontology, 2 = General Policies, 3 = User Profiles, 4 = Service Profiles, 5 = Misc. Profiles

30. Presentation Outline Introduction & Motivation
Related Work and its Limitations
Holistic Security Ontology Model
Conceptual & Service Description Layers
Policy & Application Layers
Critical Analysis
Evaluation of Conceptual Framework
Evaluation of Reasoning Model Performance
Conclusion and Future Work

31. Conclusion
There are many different security standards and they fall short of certain functionalities
Semantic Technology allows us to develop holistic models to support security interoperability, but could benefit from having more tools
An Abstract yet Explicit Model is useful for capturing stakeholders in open systems, such as V-SAT
Security Profiling advocates the understanding of security configurations and requirements amongst open services

32. Future Work
Expand the framework to support other security and semantic specifications. E.g. WSCI, ebXML.
Implement Policy based Access Control onto specific domains. E.g. KAoS, Ponder
Trust Conceptualisation & Open Interoperability

33. Thank You Contact Details: Juan Jim Tan
Department of Electronic Engineering
Queen Mary, University of London
Mile End Road, London E1 4NS
United Kingdom