A Neural Network Approach for Intrusion Detection Systems

Intrusion detection systems, alongside firewalls and gateways, represent the first line of defense against computer network attacks. There are various commercial or open source intrusion detection systems in the market; nevertheless they do not perform well in various situations including novel attacks, user activity detection, generating in some cases false positive or negative alerts. The reason behind such performance is probably due to the implementation of merely signature based checks and a high degree of dependence on human interaction. On the other hand, a neural network approach might be the right one to tackle these issues. Neural networks have already been applied successfully to solve many problems related to pattern recognition, data mining, data compression and research is still underway with regards to intrusion detection systems. Unsupervised learning and fast network convergence are some features that can be integrated into an IDS system using neural networks. The networks can be designed to process a variety of data, although there are some constraints regarding input formatting. For this reason, data encoding represents a challenging task in the integration process since it needs to be optimised for the IDS domain. This paper will discuss the integration of IDS and neural networks, including data encoding and performance issues.

Citation:
Beqiri, E.; Lee. S. W.; Draganova, C. and Palmer-Brown, D. (2010) ‘A Neural Network Approach for Intrusion Detection Systems’, 5th Conference in Advances in Computing and Technology (London, United Kingdom, 27th Jan), pp. 209 -217..