Cyber-Physical Threat Identification for Smart Systems

The constantly evolving nature of technology, the discovery of new vulnerabilities that are frequently disregarded, the variety of ways that users use their devices and perceive the threat of cyberattacks make today's Internet of Things (IoT)-based connected home challenging to secure. It's also possible for there to be a physical space breach, similar to one that occurs over the phone. Its effects are not limited to the internet. 8 In this situation, intrusion detection cannot rely only on validated models that are consistent across time and applicable to all users. 10 Introducing MAGPIE, the first intrusion detection system designed with smart homes in mind. It can automatically adjust its decision function based on anomaly classification models to a range of scenarios that may occur in a smart home, including the addition of new devices, the automation of rules, and human interaction. This is achieved by combining non-stationary multi-armed pirate reinforcement learning with a novel statistical cluster-based adjustment method. 17. MAGPIE assigns the sets of hyper parametric parameters of the underlying 18 isolation forest nonlinear anomaly classifiers based on the 19 cluster profile scores of its output. The outcomes of an experimental assessment carried out in a real home show that MAG21 PIE has remarkable accuracy because of two more innovations: it applies the models that show the highest level of precision in each scenario by taking into account data from both virtual and physical sources and by detecting human presence. Since the assessment datasets and the software are publicly available, MAGPIE may benefit from future advancements in autonomous and reinforcement learning technology. It may also be updated utilizing other data sources in the case that attacks and intelligent home settings change.