Edinburgh Napier University

  Security in computer networks is typically passive, static, and reactive. This is typically due to most networking devices being rule-based, and when updates are necessary, they are normally done manually. Ultimately, the social and hierarchical structure of an organisation should be visible within the configuration of networks. Hence, it is desirable for a distributed system to be capable of reconfiguring itself in a timely-manner to reflect changes in policy, in practices, and in the social hierarchy, such as the promotion of a member of staff, or in the face of a security threat, such as in malware propagation.

This paper builds on the concept of an automated mitigation and reconfiguration system for networked devices, and evaluates key firewall system performance tests. These could be important in defining the criteria for the success of this type of security implementation. It thus defines a range of experiments, which evaluate firewall parameters, such as number of rules, and their position in relation to performance metrics, such as CPU utilisation, bandwidth consumption, and network latency. The paper also includes tests with up to 65,000 rules, and presents results on the positions of the rules, such as on the incoming and outgoing ports, and the effect of different network throughputs.

It concludes that networks can be made more resilient, under heavy network loads and large rule sets, if rule sets are applied on the outgoing ports. It also shows evidence that configuration interfaces are the performance bottleneck for multi-agent systems that may use these to reconfigure network equipments dynamically.