Latent Dirichlet Allocation (LDA) for Anomaly Detection in Avionics Networks

Abstract

Latent Dirichlet Allocation (LDA) and Variational Inference are applied in near real-time to detect anomalies in ground vehicle network traffic for a ground vehicle network. The technical approach, that utilizes the Natural Language Processing (NLP) technique to detect potential malicious attacks and network configuration issues, is described and the results of a proof of concept implementation are provided. Potential use cases for applying the technique in the aircraft and avionics domain are provided.