A Trusted Hybrid Learning Approach to Secure Edge Computing

Abstract

Securing edge computing has drawn much attention due to the vital role of edge computing in Fifth Generation (5G) wireless networks. Artificial Intelligence (AI) has been adopted to protect networks against attackers targeting the connected edge devices or the wireless channel. However, the proposed detection mechanisms could generate a high false detection rate, especially against unknown attacks defined as zero-day threats. Thereby, we propose and conceive a new hybrid learning security framework that combines the expertise of security experts and the strength of machine learning to protect the edge computing network from known and unknown attacks, while minimizing the false detection rate. Moreover, to further decrease the number of false detections, a cyber security mechanism based on a Stackelberg game is used by the hybrid learning security engine (activated at each edge server) to assess the detection decisions provided by the neighboring security engines.