Industrial IoT systems are constantly increasing security to mitigate failures and unallowed access. Nowadays, enterprise systems can handle large amounts of sensitive information that can be received from hundreds of...
Industrial IoT systems are constantly increasing security to mitigate failures and unallowed access. Nowadays, enterprise systems can handle large amounts of sensitive information that can be received from hundreds of tiny devices such as sensors, fog devices, gateways, etc. However, the lower processing capacity associated with tiny devices due to manufacturing decisions has contributed to their lack of defensive measures and hardware security support. Those devices performance usually restricts advanced software security usage and results in several vulnerabilities in the device’s authentication. RFID tags have been used in the IoT industry to create a variety of applications for identification, authentication, or even virtual proof of reality, but RFID tags are still susceptible to serious attacks since fake RFID devices or other attacks are trying to find ways of bypassing the security. In this paper, 5 (five) machine learning algorithms have been used to classify a dataset of read-time operations from blocks in Mifare RFID and fake tags, aiming to identify different behaviors between dataset’s variables, which can be used for a Physical Unclonable Functions (PUF). According to our results, K-Nearest Neighbor (KNN) offered the best performance for the considered problem. The KNN obtained a root mean squared error equal to 0.058 and an mean absolute error equal to 0.003. The obtained results indicate that operation times have variables inherent to the manufacturing process and this can be useful for future identity management schemes.
暂无评论