Antiference: New Concept for Evolutive Mitigation of RFI to GNSS

Abstract

The past decade has shown a growing awareness of the dangers of intentional interference (especially jamming and spoofing) with GNSS signals. The Antiference project uses reconfigurable digital signal processing methods in the detection, classification, and mitigation of interference by employing machine learning techniques. The ML-based jamming classifier uses distinctive features of spectrograms for the differentiation of various jamming attacks. A residual neural net is used to map the spectrograms to the different jamming types. It relies on a fingerprinting architecture. Fingerprints summarize the characteristics of all the incoming signals, which are stored in and matched to a database of previously encountered interference types. To validate the implemented functionalities, a developed test-bed runs test scenarios and benchmarks the results against two state-of-the-art COTS receivers with interference mitigation capabilities.

Keywords

• machine learning
• jamming
• spoofing
• ResNet50
• COTS receivers
• interference detection and mitigation