Adaptive Neuro-Fuzzy Inference System-based Lightweight Intrusion Detection System for UAVs

Unmanned aerial vehicles (UAVs) are widely utilized in myriad domains due to their low infrastructure cost and flexibility in deployment. Hostile and unsafe networking environments can make UAVs vulnerable to various attacks. Intrusion detection systems (IDSs) have been developed to detect such attacks. However, conventional data-driven IDSs can be architecturally complex and computationally intensive for resource-constrained small UAVs. In this work, we propose a lightweight IDS for UAVs leveraging an adaptive neuro-fuzzy inference system (ANFIS) that combines artificial neural networks (ANNs) and fuzzy deduction frameworks. Due to the simplistic membership and rule-based classification capabilities of ANFIS, our proposed IDS is lightweight and perfectly suitable for small UAVs. We evaluate the ANFIS-IDS’s effectiveness by comparing its performance to conventional data-driven classification models. In particular, we contrast the proposed IDS with a traditional novelty-based IDS for UAV sensor attacks. We further compare their deployment in a hardware-emulated UAV testbed, assessing the proposed model’s lightweight nature.