Traffic Load Learning Towards Early Detection of Intrusion in Industrial mMTC Networks

The fifth generation network offers a powerful infrastructure for Industrial internet of things (IoT) in that its support on massive machine-type communications (mMTC). Recently, grant-free access has been recognized as the promising new access approach for mMTC; however, it also introduces the potential risk. To timely discover intrusion, in this article, we propose a learning network by extracting traffic load information from the states (success, collision, and idle) of access resources observed at media access control and physical layers. In particular, our learning network consists of three concatenated function modules, i.e., traffic load estimation, traffic load prediction, and intrusion identification. Moreover, our proposed learning network is able to identify two types of intrusion, i.e., false data dissemination and congestion attack. Simulation results indicate that the proposed scheme can effectively capture the number of active devices, provide reasonable prediction by using history records, and eventually, achieve more accurate detection compared with baseline approaches.