More and more portable intelligent devices are connected to the Internet in recent years. A way to effectively use the isolated cyber data without involving privacy and realize the cyber intrusion anomaly detection on the portable intelligent devices with relatively limited hardware storage resources and computing power is worth exploring. In this paper, we propose a framework of federated anomaly detection, which enables the device effectively detect the anomaly by sharing the parameters of the federated model in a fully distributed fashion. We formulate the model training problem as a distributed robust optimization problem and subsequently devise an efficient algorithm for it. Experimental studies have also been carried out to reveal the superior performance of the proposed framework and underscore the significant benefits of federated anomaly detection.
This work is supported by the MSIT (Ministry of Science and ICT), Korea, under the ITRC (Information Technology Research Center) support program (IITP-2023-2018-0-01431) supervised by the IITP (Institute for Information & Communications Technology Planning & Evaluation).
