Cognitive Immunity Support for Large-Scale Autonomic Software Systems

1
Cognitive Immunity Support for Large-Scale
Autonomic Software Systems

• David Lamb
• D.Lamb_at_2005.ljmu.ac.uk
• Room 608, ext. 2280
• http//www.staff.ljmu.ac.uk/cmpdlamb
• Supervisors Dr. Dhiya Al-Jumeily, Prof. A
  Taleb-Bendiab
• School of CMS, Liverpool John Moores University

2
Overview

• Background Situating the problem
• Research Literature Review
• Motivation Why this line of research
• Current Research overview
• Research Objectives
• Identified Problem Areas
• Future Work plans for the future

3
Background

• Setting the scene
• Who else is doing similar work?
• Overview of literature review
• Cognitive Immunity
• Artificial Immune Systems
• Cognitive Systems
• Machine Learning, etc
• Complex Networks / Graph Theory

4
Background Cognitive Immunity

• Influential/Related work
• Notion of CI discussed as one approach in DARPA
  Self-Regenerative Systems research programme
• DARPA define CI systems as those
• Capable of accurately diagnosing root causes of
  system problems
• Capable of taking effective corrective action
  appropriate to problem diagnoses

5
Background AIS

• Artificial Immune Systems
• Many aspects influenced by Biologically-Inspired
  Computing
• Bottom-up approach
• Creates complex behaviour from simple
  interactions
• Therefore, may scale well to manage complex
  computer systems
• Some examples
• Biological Immune System
• Artificial Immune Systems
• Self/Non-self discrimination Pattern
  Recognition
• Danger Theory
• Evolution GAs
• Emergence Ants, Swarms, etc

6
Background Cognitive Systems

• ALCS (Anticipatory Learning Classifier Systems)
• ALP Anticipatory Learning Process
• Observes environment
• Generates specialised rules that describes the
  observed behaviour
• Of the form (Condition ? Action ? Effect)
• GGM Genetic Generalisation Mechanism
• Uses genetic selection mechanism to generalise
• Keeps rule set correct and compact

7
Background Machine Learning

• Machine Learning
• Novelty Detection
• Statistics clustering of types
• Neural Networks trained networks as classifiers
• SOMs self-organising classifiers
• Chance Discovery
• Change-based KB, Dialogue approach, Key graphs
• Reinforcement Learning

8
Background Graph Theory

• Graph Theory
• May provide a method to understand complex
  systems organisation
• Identifies nodes and connections
• Understanding which nodes form hubs
• Complex Networks
• E.g. Small world and scale-free networks
• Scale Free robust in random failures

9
Motivation

• What makes this research worthwhile?
• Brief overview of
• Static vs. Dynamic software design
• Benefits of Dynamic/Evolving systems
• Problems involved in dynamic system design

10
Motivation Software Design

• Traditional Static System Design Methods
• Well understood
• Limitations
• Resistant to change
• Inadequate for modelling complex systems
• Design Methods for Complex, Large-Scale, Dynamic
  Systems
• Would overcome some limitations
• New Problems and Challenges

11
Motivation Dynamic Systems

• Dynamic System Design
• Should allow the system to
• Evolve at runtime
• Allows optimal (re) configuration and
  organisation
• Respond to changing environments
• Resist threats
• However, brings its own problems
• How to design it?
• How to best implement it?
• How to support it?

12
Current Research

• What have I done?
• Literature Review
• Prepared Research Proposal
• What has that achieved?
• Identified Research Objectives
• Further Research Problems

13
Current Research Literature Review

• Literature Review
• Machine Learning Techniques
• Artificial Immune Systems
• Cognitive Immunity
• Cognitive Systems
• Complex Networks / Graph Theory

14
Current Research Objectives

• Literature review led to research proposal,
  identifying the following objectives
• Further Literature Review of Cognitive Systems
• Creation of a programming model and framework for
  Evolving, Self-Healing systems that demonstrate
  Cognitive Immunity
• Understand Requirements of this approach
• How to develop and support this approach
• How to apply this approach

15
Current Research Problems

• Identified Research Problems relevant to creating
  a system capable of Cognitive Immunity
• Lack of formalised programming models
• Adaptation to Environment
• Environmental Sensing
• Plan Generation
• Plan Enactment
• Benevolent System Observation
• Tuning, Improvement, Optimisation

16
Future Work

• Further Literature Review
• More on ALCS, including a prototype
  implementation
• Research other Cognitive-type systems
• Graph Theory, and approaches to Complex Networks
• Other suitable models for self-organising systems
• Definition of Requirements Model
• leading to the creation of the related
  programming model
• Further development and generalisation of the
  model

17
Thank you for listening!

• Any Questions?