Contextual Insight: Detecting Abnormal Device Behaviors in IoT Systems
Jian Ge, Jianwu Rui, Hengtai Ma, Bin Li, Yeping He · 2024
The proliferation of collaborative technologies and the Internet of Things (IoT) has led to a greater diversity of devices. However, it also increased their vulnerability to potential attacks. The inherent heterogeneity and complexity of IoT devices make it difficult to accurately detect abnormal behavior, leaving them open to manipulation by attackers. To address this issue, we propose Talos, an innovative solution to detect abnormal device behavior in IoT systems. Talos employs advanced deep neural networks to analyze and comprehend historical event sequences, enabling it to accurately model the normal behaviors of IoT devices and detect any anomalies through reconstruction error. We evaluate our prototype on a real-world testbed and test it for three different anomaly cases. The experimental results demonstrate that Talos achieved high precision and recall, achieving an average F1-Score of 99.73% while raising only a few false alarms.