Unsupervised Machine Learning for Malicious Network Activities

Abstract

Increase in cybercrimes in the last few years has dramatically increased the need for the network intrusion detection and subsequently its mitigation. Several classified intrusion detection systems (IDS) are currently in use but increase in dynamics of the cyber invasion hunts for more adaptive and intelligent model. This research covers the critical analysis and comparison of the Machine Learning (ML) network intrusion detection techniques, their use cases and proposes an unsupervised and fast ML implementation model for intrusion detection. The proposed model works on anomaly-based detection. ELK stack (Elasticsearch, Logstash and Kibana) has been used for unsupervised implementation model for exodus DNS requests in a wired network