On Detecting IoT Power Signature Anomalies using Hidden Markov Model (HMM)

The internet of thing (IoT) has become a widespread phenomenon that has resulted in an exponential growth in data. The increase in the number of devices connected to the internet has raised a lot of security issues and privacy concerns. Users are susceptible to many forms of cyber-attacks; which could affect the functionality of their devices and their power signature. In this paper, we are proposing the usage of Hidden Markov Model (HMM) to model the power consumption patterns of IoT sensor nodes. Using such model, we were able to detect anomalies in the system's behavior and thus, identifying attacks and classifying them. Using the training data, a detection threshold is computed to differentiate between normal and anomalous behavior. A series of tests were performed on the algorithm using false data injection and the effect of varying HMM parameters is studied. Finally, alterations to the IoT sensor node's functionality are made to mimic the effect of an attack. The method was found to be successful in identifying attacks, trojans and malfunctions as well as, detecting the type of change that has occurred in the system's performance.