Unmanned Aerial Vehicle (UAV) Cyber Attacks: Anomaly Detection Using Neural Network-Based Intrusion Detection
Abstract
The use of unmanned aerial vehicles (UAVs) in civilian, commercial, and military applications has grown and continues to grow. An emerging area of interest for UAVs is for several to form flying ad-hoc networks (FANETs). In both military and non-military settings, FANETs can be used for tasks such as extending communication networks, managing crops, or conducting surveillance. However, with the increased use of UAVs in the national air space, it is important to prevent or mitigate cyber-attacks such as spoofing or anomaly insertion. Neural network-based intrusion detection techniques provide a way to potentially identify anomalous data that indicates a cyberattack. Although well-trained neural networks can classify data and make predictions with a high degree of accuracy, they are not infallible. This research explored the potential of using hierarchical clustering techniques to identify anomalous data points within neural network hidden layer activations, and identifying the nodes in which the activations adversely impacted the results.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 31, 2022
- Accession Number
- AD1230860
Entities
People
- Sherida T. Jacob
Organizations
- Naval Undersea Warfare Center