Applying big data based deep learning system to intrusion detection

• DOI: 10.26599/bdma.2020.9020003
• Publication Date: 2020-07-16T19:56:37Z
• Authors: Wei Zhong; Ning Yu; Chunyu Ai
• Keywords: Interconnectivity; Payload (computing)

With vast amounts of data being generated daily and the ever increasing interconnectivity world's internet infrastructures, a machine learning based Intrusion Detection Systems (IDS) has become vital component to protect our economic national security. Previous shallow deep strategies adopt single model approach for intrusion detection. The may experience problems understand increasingly complicated distribution patterns. Particularly, not be effective capture unique patterns from intrusive attacks having small number samples. In order further enhance performance IDS, we propose Big Data Hierarchical Deep Learning System (BDHDLS). BDHDLS utilizes behavioral features content both network traffic characteristics information stored in payload. Each concentrates its efforts learn one cluster. This strategy can increase detection rate as compared previous approaches. Based on parallel training big techniques, construction time is reduced substantially when multiple machines are deployed.