Optimizing the Scalability of Network Intrusion Detection Systems Using Mobile Agents

Modern Intrusion Detection Systems (IDSs) are distributed real-time systems that detect unauthorized use or attacks upon an organization's network and/or hosts. The components of most distributed IDSs are arranged in a hierarchical tree structure, where the sensor nodes pass information to the analyzer nodes. Optimal placement of the analyzer nodes results in an improved response time for the IDS, and isolation of attacks within the IDS network. Since the network topology and workload are constantly changing, we are able to maintain near-optimal placement of the analyzer nodes by instantiating them as mobile agents. The analyzer nodes may then relocate, reproduce or be deleted as necessary. Such flexibility improves the response times and the stability of an IDS. The movement of the analyzer nodes also offers some protection against denial-of-service attacks, since secure analyzer nodes will be relocated to take over some of the functionality of the host under attack.