Securing the Internet of Things Through Device Identification via Network Traffic Analysis

The widespread adoption of the Internet of Things (IoT) has led to considerable security concerns, particularly in device identification. To tackle this challenge, we have devised a method that scrutinizes the data transmitted by IoT devices over the network, delving into detailed behavioral patterns that differentiate one device from another. Acknowledging the similarities in device behaviors, our approach concentrates on discerning and categorizing them based on their unique characteristics. This paper introduces an approach to identifying IoT devices by integrating probabilistic fusion and convolutional neural networks (CNNs). Our method involves segmenting input data into overlapping subsequences and using a CNN to independently classify each subsequence, enabling the detection of intricate patterns. The predictions are combined using an averaging method, leading to improvements. Specifically, the MFT dataset F1-score increased by 9.3%, the PC dataset by 2.46%, and the BMSAQ dataset by 24.6%. Empirical validation demonstrates that our model surpasses existing methods, offering a practical and efficient solution for real-world IoT applications. This research underscores the potential of advanced machine learning techniques in fortifying and safeguarding IoT ecosystems.