On one hand, Network Intrusion Detection Systems (NIDS) that are based on machine learning sometimes demand important domain expertise and experimentation. That makes performance be suboptimal in complex network environments. On other hand deep learning although its power can hardly deal with imbalanced data. That results in a bias towards normal traffic and reduced effectiveness in detecting rare attacks.

To overcome this issue, researchers propose a method that combines contrastive learning and Bayesian Gaussian Mixture Model (BGMM). With this method, they can automatically learn similarities within normal traffic and the differences between normal and malicious traffic, so generate distinguishable and strong feature representations.

Learn more: https://cybersecurity.springeropen.com/articles/10.1186/s42400-025-00364-7

Authors: Liyou Liu & Ming Xu