Intrusion Detection Using Incremental Learning from Streaming Imbalanced Data

Most of the network habitats retain on facing an ever increasing number of security threats. In early times, firewalls are used as a security examines point in the network environment. Recently the use of Intrusion Detection System (IDS) has greatly increased due to its more constructive and robust working than firewall. An IDS refers to the process of constantly observing the incoming and outgoing traffic of a network in order to diagnose suspicious behavior. In real scenario most of the environments are dynamic in nature, which leads to the problem of concept drift, is perturbed with learning from data whose statistical attribute change over time. Concept drift is impenetrable if the dataset is class-imbalanced. In this review paper, study of IDS along with different approaches of incremental learning is carried out. From this study, by applying voting rule to incremental learning a new approach is proposed. Further, the comparison between existing Fuzzy rule method and proposed approach is done.