Transforming voluminous data flow into continuous connection vectors for IDS

Abstract

Handling massive data flows in the online intrusion detection systems is considered one of the major challenges in computer network security. This paper presents a novel method that overcomes this challenge using data mining and statistical techniques. The main contributions of the proposed method are: 1 capturing network traffic and hosts' activities using an intelligent aggregator; 2 introducing a queuing concept of dynamic window size; 3 an improved correlation method between network packets and hosts' events; 4 exporting continuous connection vectors based on certain features set. The proposed method has been evaluated using offline, synthetic, and realistic data flow. Moreover, it is compared to other competent methods. We have examined the plausibility and scalability of the constructed connection vectors by using them in evaluating intrusion detection models.