Light Auditor: Power Measurement can tell Private Data Leakage through IoT Covert Channels

Despite many conveniences of using IoT devices, they have suffered from various attacks due to their weak security. Besides well-known botnet attacks, IoT devices are vulnerable to recent covert-channel attacks. However, no study to date has considered these IoT covert-channel attacks. Among these attacks, researchers have demonstrated exfiltrating users' private data by exploiting the smart bulb's capability of infrared emission. In this paper, we propose a power-auditing-based system that defends the data exfiltration attack on the smart bulb as a case study. We first implement this infrared-based attack in a lab environment. With a newly-collected power consumption dataset, we pre-process the data and transform them into two-dimensional images through Continous Wavelet Transformation (CWT). Next, we design a two-dimensional convolutional neural network (2D-CNN) model to identify the CWT images generated by malicious behavior. Our experiment results show that the proposed design is efficient in identifying infrared-based anomalies: 1) With much fewer parameters than transfer-learning classifiers, it achieves an accuracy of 88% in identifying the attacks, including unseen patterns. The results are similarly accurate as the sophisticated transfer-learning CNNs, such as AlexNet and GoogLeNet; 2) We validate that our system can classify the CWT images in real time.