Adaptive Systems Lecture 6: Adaptation Mechanisms (part 3)

1
Adaptive SystemsLecture 6 Adaptation
Mechanisms (part 3)

• Dr Giovanna Di Marzo Serugendo
• Department of Computer Science
• and Information Systems
• Birkbeck College, University of London
• Email dimarzo_at_dcs.bbk.ac.uk
• Web Page http//www.dcs.bbk.ac.uk/dimarzo

2
Lectures 4/5 Review

• Part 1 Adaptation Mechanisms (Lecture 4)
• Stigmergy
• Schools of fish, flocks of birds
• Autopoiesis
• Decrease of Entropy
• Trust
• Holarchy
• Part 2 Adaptation Mechanisms (Lecture 5)
• Economic Based Mechanisms
• Reinforcement Learning

3
Lecture 6 Overview

• Adaptation Mechanisms
• Non-naturally inspired
• Chains of authentication
• Tags
• Self-management

4
Target Applications

• Networking
• Grid systems
• P2P Systems
• Security - Protocols
• Autonomic Computing
• Self-managing systems
• Ambient Intelligence
• Seamless delivery of user services and
  applications
• Based on ubiquitous and pervasive computing and
  communication embedded in the users environment.

5
Chains of Authentication

• Public-key cryptography
• Encryption with public key ensures
  confidentiality (only private key owner can
  decrypt)
• Encryption with private key ensures authenticity
  (of digital signatures)
• Problem ensuring authenticity of public key
• Usual solution use a third party certificate
  authority (central solution)
• Third party authenticates the validity of the
  pair (public key, private key)
• MANET Mobile Ad Hoc Networks
• No fixed infrastructure
• Management tasks performed by nodes
• No central repository of certificates

6
Chains of Authentication

• Self-Organised Public-Key Management for MANETs
  Capkun 03
• Authentication of public key certificates in
  mobile ad hoc networks
• Users issue public-key certificates to other
  users
• Public key certificate of (identity)
  authenticity
• Certificate has a limited validity period (but
  can be updated extended)
• Certificates are stored and distributed by the
  nodes and each node maintains a local certificate
  repository
• Key authentication is performed via chains of
  public-key certificates
• Nodes locally maintain their own list of valid
  public key certificates
• Nodes locally check certificate authenticity and
  validity
• No central authority, no central repository
• Correctness and checks are performed locally

7
Chains of authentication
Nodes (Public Keys)
y
CertificatesId of w certified by u
u
w
w (Kw)
u (Ku)
z
(Idw)PrKu
x
v (Kv)
8
Chains of Authentication

• Construction
• Each user creates its own public/private key
  pairs
• Third party issues public-key certificates
• Certificates exchange (with neighbours)
• Nodes construct local repositories of
  certificates
• Permanent update of repositories through node
  communication (revocation of certificates)

9
Chains of Authentication

• Scenario
• u wants to authenticate public key of v
• Merging of certificate repositories of u and v
• Path determination
• Paths of certificates (from u to v)
• u finds a path of certificates
• u verifies
• First certificate of the chain with own public
  key (Ku) (valid correct)
• Second certificate of the chain with key obtained
  with first certificate
• until last certificate for v, which contains Kv.

10
Chains of authentication
Nodes (Public Keys)
y
CertificatesId of w certified by u
u
w
w (Kw)
u (Ku)
z
(Idw)PrKu
x
z guarantees id of v y guarantees id of
z x guarantees id of y w guarantees id of
x u guarantees id of w
v (Kv)
11
Adaptation

• MANET Nodes join and leave frequently
• Certificates have an expiry time
• Local repositories are regularly updated
• Paths allow to adapt to new (current) status of
  valid certificates
• No central entity

12
Tag-Based Models

• Tag-Based Models Hales 04
•  Markings attached to individuals (agents) and
  observable by others 
• Markings bit strings, list of neighbours (P2P)
• Agents interact with others similar to them
  (similar tag)
• Agents with similar tags are closer to each
  other
• Applications
• Prisoners dilemma
• Improving cooperation in P2P networks
• Peer-to-peer (P2P) networks
• Networks of peer nodes (no client/server
  relations)
• Ad hoc connections among nodes
• Usually for file sharing Kazaa, Gnutella,
  Bittorrent

13
Tag-Based Models

• SLAC Algorithm Hales 05
• Selfish Link and behaviour Adaptation to produce
  Cooperation
• Tag-based
• Assumption peer nodes can change their behaviour
• Copy and re-wire rule based on a utility
  function Mutation
• Copy all links of another node
• Add a link to that node and drop all previous
  links
• Copy behaviour of that node
• Tag utility function links
• Applications
• Self-organise P2P networks into tribes
  (specialised groups)
• Nodes with high utility tend to replace those
  with low utility
• Cooperating groups are reinforced, while selfish
  groups disappear

14
Tag-Based Models

• Algorithm
• Nodes store list of neighbours
• Nodes compute their own utility (e.g. number of
  files downloaded)
• Periodically
• Each node x compares its own performance Ux
  (utility) against that of a randomly chosen node
  y, Uy
• If Ux lt Uy, then
• x drops its current links and copies those of y
  AND
• x adds a link to y
• Mutation x drops all links, adds one link to a
  randomly chosen node
• Effect
• Nodes tend to copy themselves in order to
  increase their utility
• Groups of nodes with similar tasks (favours
  altruistic behaviour instead of selfish behaviour)

15
Examples

• Scenario
• Nodes
• Skill (numbers 1 to 5)
• Altruism flag 0,1
• Utility function u
• Different kinds of Resources (1 to 5)
• Need to be processed by nodes with the
  corresponding skill
• Node n receives a job
• Node n has right skill un un 1
• Node n does not have right skill, then passes job
  to immediate neighbour k that has right skill
  altruist flag 1
• un un 1
• uk uk 0.25
• Nodes can change skill
• Utility measure
• Percentage of submitted jobs completed
• If high percentage of success then network is
  cooperative
• A job is completed if passed to a neighbour

16
Examples

• Simulation
• Initialisation
• All nodes have randomly selected skills (numbers
  1 to 5)
• Random links among nodes
• Utility measure
• Percentage of submitted jobs completed
• Apply SLAC Algorithm
• Algorithm runs over several cycles
• Specialised groups emerge
• http//www.cs.unibo.it/pub/TR/UBLCS/2005/2005-13.p
  df

17
Adaptation

• Specialised tasks done collaboratively
• Specialisation of nodes (re-wire)
• Grouping of nodes for similar tasks (tribes)
• Increased collaboration

18
Autonomic Computing

• Human nervous system
• Governs vital functions (heart, digestion, body
  temperature, etc.)?
• Not consciously controlled
• Conscious brain can concentrate on other tasks
• Autonomic Computing System
• Act as human nervous system
• Provide vital functions to distributed system
• Underlying system not conscious of this
• http//www-03.ibm.com/autonomic/

19
Autonomic Computing

• Self-Configuration (installation, configuration,
  integration)?
• Automated configuration of components and
  systems follow high-level policies. Rest of
  System adjusts automatically and seamlessly
  Kephart03
• Self-Optimisation (parameters)?
• Components and systems continually seek
  opportunities to improve their own performance
  and efficiency Kephart03

20
Autonomic Computing

• Self-Healing (error detection, diagnostic,
  repair)?
• System automatically detects, diagnoses, and
  repairs localized software and hardware problems
  Kephart03
• Self-protection (detection and response to
  attacks)?
• System automatically defends against malicious
  attacks or cascading failures. It uses early
  warning to anticipate and prevent system wide
  failures Kephart03

21
Autonomic Computing

• Autonomic Computing / Self-Management
• Self-configuration (installation, configuration,
  integration)
• Self-repair (error detection, diagnostic, repair)
• Self-optimisation (parameters)
• Self-protection (detection and response to
  attacks)
• Techniques for Self-Management (not bio-inspired)
  
• Separation
• Interface and Behavioural Requirements of
  autonomic components / Interactions among
  components White 04
• Computation Behaviour / Composition Run-time
  injected interaction rules Liu 04

22
Dynamic Composition of Autonomic Components

• Autonomic Component
• Communication ports (functional, control,
  operational)
• Behavioural rules (functional behaviour)
• Interaction rules (among components)
• Dynamic composition of autonomic components
• Organisation of components
• Interactions among components
• Dynamic replacement, adding/deleting components
• Scenario Liu 04
• Interactions triggered by a component or an event
• Composition agent
• Defines set of interaction rules
• Peer Rule Agents responsible to execute
  interaction rules
• Configuration of components
• Establishment of interactions
• Actually build interactions in a decentralised way

23
Adaptation

• Summary
• Seamless integration of new available services
• Robustness against faulty services
• Interoperability among independent business
  applications
• Self-management

24
Reading

• Capkun 03 S. Capkun, L. Buttyan, and J.P.
  Hubaux. Self-Organized Public-Key Management for
  Mobile Ad-Hoc Networks. Transactions on Mobile
  Computing, 2003.
• Hales 04 D. Hales. Self-Organising Open and
  Collaborative P2P Societies From Tags to
  Networks. In S. Brueckner et al. editors,
  Engineering Self-Organising Applications
  (ESOA'00), LNAI 3464, Germany, 2005.
  Springer-Verlag.
• Hales 05 D. Hales. Choose Your Tribe! Evolution
  at the Next Level in a Peer-to-Peer Network. In
  S. Brueckner et al. editors, Engineering
  Self-Organising Applications (ESOA'05), LNAI to
  appear, Germany, 2005. Springer-Verlag.
• Liu 04 H. Liu, M. Parashar, S. Hariri. A
  component-based programming model for autonomic
  applications. First international conference on
  autonomic computing (ICAC04) pp. 10-17.
• White 04 S. White, J. Hanson, I. Whalley, D.
  Chess, J. Kephart. An architectural Approach to
  Autonomic Computing. First international
  conference on autonomic computing (ICAC04) pp.
  2-9