Semi-Supervised Transformation and Deep Embedding-Based Anomaly Identification for Agricultural Internet of Things
Xiang Yin, Letian Wang, Weikuan Jia, Chengqian Jin · IEEE Sensors Journal · 2020
This study aims at resolving the problem of anomaly identification for network flows in Internet of Things (IoT), and propose a novel netflow detection approach. Accurate recognition of anomaly is of fundamental importance to numerous network activities including cyberattack monitoring and intrusion prevention. In the past decades, a plethora of techniques have been presented to address this issue. Furthermore, the difference between the anomalous and normal networking activities is trivial in common. Therefore, the machine learning-based algorithms might be incapable of dealing with it. Bearing this in mind, a transformation-based pipeline is introduced to reveal the inconspicuous anomalies. The main idea behind our scheme is to apply various transformations upon data samples would be beneficial to accurate anomaly detection. The comparison experimental results indicate that our work could outperform the accuracy of state-of-the-art techniques.