Toward a realistic Intrusion Detection System dedicated to smart-home environments
Olivier Lourme, Michaël Hauspie · 2021
The Internet of Things (IoT) is a recent and ever growing model in data processing. In terms of security, the heterogeneity and specificity of protocol stacks, the variety and low resources of objects, combined with commercial pressures, lead to less mature and less robust solutions than those available in traditional Information Technologies, hence providing a new attack surface, often exploited.In this context, the development of Intrusion Detection Systems (IDS) dedicated to IoT is an abundant research field. Inside it, many works claim to tackle the peculiar and promising IoT smart-home segment. Sadly, a lot of them do not deal with its specific characteristics compared to other IoT fields: 1) its multiple protocol stacks in a small volume, 2) its reinforced economic stress and 3) the lack of technical skills from users waiting primarily for things to work without any hassle.In this paper, we propose a smart-home IDS design, driven by the aforementioned characteristics of smart-home environments (technical, economic and human). For example, acquisitions and demodulations of signals should be performed by low-cost multiprotocols dongles, not needing any calibration. The anomaly detection algorithm, implemented in an updatable centralized host, should be taken among the unsupervised learning methods, less expensive and simpler than supervised alternatives. We believe that this new holistic approach, if it meets satisfactory performance metrics, may contribute significantly to a wide adoption of IDS in smart-home environments.