Unsupervised Learning for Network Intrusion Detection | Nandi Leslie | Raytheon | WiDS 2022
About This Video
Nandi Leslie, Engineering Fellow at Raytheon Technologies, presents a Technical Vision Talk at the WiDS Worldwide conference.
For nearly 40 years, computer scientists and engineers have been concerned with the problem of monitoring networks for unauthorized activities. More recently, anomaly-based intrusion detection systems have been developed to protect enterprise and mobile networks from such attacks. Nonetheless, in-vehicle networks remain vulnerable to a variety of remote attacks that erode information confidentiality, availability, and integrity.
In this talk, Nandi develops an ensemble hierarchical agglomerative clustering (E-HAC) algorithm for detecting remote attacks on the CAN bus. E-HAC is an ensemble learning approach over multiple clustering algorithms with different linkages and pairwise distances between observations. In addition, she presents prediction performance results for a dataset consisting of CAN bus and remote attack network traffic to demonstrate the effectiveness of this E-HAC algorithm.