Classification and Anomaly Detection of Network Traffic at the edge using Transfer Learning

Anomaly network traffic threat detection has been used in many companies to detect threats. Usually, it is deployed on-premises for efficiency and privacy reasons. Anomaly detection system usually consists of a machine learning algorithm that needs tons of time to train and test. Because each company may have different network setup, it is hard to build one system that suits for everyone. So, we need to customize the system for each client. Nevertheless, it is very inefficient to retrain the algorithm from zero for each company. Therefore, we hope to prove that by using transfer learning, we can move the knowledge from an existing detection system to a new network environment without retraining the whole algorithm.

Faculty Supervisor:

Ali Dehghantanha

Student:

Chenxingyu Chen

Partner:

eSentire Inc.

Discipline:

Computer science

Sector:

Information and cultural industries

University:

University of Guelph