ABSTRACT
An increasing number of edge devices store and process sensitive user data, presenting an attractive target for attackers. This trend of data storage and processing at the edge is expected to continue. As secure devices are integrated into new systems with increased device operation times, exposure to environmental stress also increases significantly. Especially, for standalone micro-Edge devices the relevance of this is increasing. Enhanced protection mechanisms are required and AI-based approaches are promising candidates.
In this contribution, we examine the requirements for such mechanisms and the sensing capabilities of state-of-the-art secure devices. Based on these capabilities and attack models, a dataset for training and validation is generated. Considering the requirements and the available dataset, a selection of applicable algorithms is defined. The selected algorithms are evaluated and compared based on the obtained results and computational loads, as the basis for future work.
