Eureka! Institutional Repository

Embedded devices, such as programmable logic controllers (PLC) and Internet-of Things (IoT) devices are becoming targets of malware attacks with increasing frequency and catastrophic results. Physical side channel analysis is one way to monitor the device without accessing its software, thus causing no resource overhead to the device. In this thesis we presented an alternative way of using side channel analysis for detecting anomalies in embedded devices during code execution. We used power consumption side channel signals to design an intrusion detection system based on convolutional neural networks. We used the proper equipment to capture signals representing different paths of the execution code. Considering these paths as classes we fed a convolutional neural network we created in order to train it to predict when there is an intrusion leading to an code execution abnormality. There is a previously published relative work on intrusion detection based on EM side channel analysis. In this thesis we focus on power side channel signals.