Communications of the ACM
Artficial Immune Systems and Their Applications
Artficial Immune Systems and Their Applications
WET-ICE '96 Proceedings of the 5th International Workshops on Enabling Technologies: Infrastructure for Collaborative Enterprises (WET ICE'96)
Self-Nonself Discrimination in a Computer
SP '94 Proceedings of the 1994 IEEE Symposium on Security and Privacy
Security in an autonomic computing environment
IBM Systems Journal
Online Monitoring and Analysis for Self-Protection against Network Attacks
ICAC '04 Proceedings of the First International Conference on Autonomic Computing
Moving up the information food chain: deploying softbots on the world wide web
AAAI'96 Proceedings of the thirteenth national conference on Artificial intelligence - Volume 2
Hi-index | 0.01 |
The immune system response of the vertebrates demonstrates an extremely adaptive and resilient defensive capability against a broad spectrum of pathogenic attacks. The field of artificial immune systems (AISs) aims to replicate this capability in the digital environment. In particular, we would like to understand adaptive survivability and defence in large-scale computing networks. In this paper we discuss some of the background concepts to AIS and focus on one specific aspect required to achieve a digital immune system, i.e. the social dynamics of competitive and co-operative defence. In particular, the ability of an information network to maintain itself in the face of continuous perturbation raises more complex issues related to system metabolism and network topology. In order to investigate these processes a multi-agent simulation model has been developed that demonstrates a self-organising group formation capability and a collective immune response. In this model each agent is susceptible to viral infections passed between the agents and has local sensors and a complex metabolic state that reflects its current health. We then introduced an artificial immune system to each agent that allowed learned 'antibody' solutions to be exchanged between the agents within a social group. The health of a co-operative group was observed to improve by over 90%, relative to isolated agents or non-cooperative groups. The specific solutions advocated are therefore to utilise distributed defence mechanisms and the monitoring of metabolic processes to detect intrusions. Finally, the paper considers the problem of how we might utilise such knowledge to develop greater security and robustness in real-world networks using distributed agent systems.