Network Traffic Attention Features Fusion for Unauthorized IoT Devices Detection
The demand for using IoT devices is increasing because of its usefulness, IoT devices are present in all areas of life, and it makes life easier and more comfortable. Moreover, many companies and households are tending to increase in the use of IoT devices to turn on/off home, school, office into smart home, smart school, smart office and so on. Although, IoT devices are very useful, but they have major security problems because they are not interested, leading to a large number of security holes and creating opportunities for hackers to attack the systems. In this paper, we develop an attention features fusion (AFF) method for the purpose of classifying and detecting IoT devices that are not in the pre-registered white list. Our recommended technique finds the features with the best classifier ability, then converts them into images and uses ensemble learning based on the baseline models which are deep learning models to detect and classify. Our proposed solution helps to detect strange devices that are exchanging data in the IoT network, reducing the risk of the system being attacked by hackers. According to the experimental results, our method achieves very good results when we detect ten IoT devices with up to 97.85\% accuracy.