The rapid proliferation of the Internet of Things (IoT) has garnered significant attention and is poised to profoundly transform daily life. IoT devices offer a wide range of applications, spanning from smart homes to sophisticated business networks and medical care systems. However, this widespread adoption also exposes IoT devices to potential security threats and malicious attacks, posing a persistent risk to the safety of sensitive data they handle. The consequences of security breaches extend beyond individual victims to affect the broader global community. In response to this challenge, machine learning (ML) techniques have emerged as promising tools for enhancing IoT device security. This paper proposes a novel unsupervised ML-based framework designed to efficiently detect anomalous signals, and thereby enabling active Reconfigurable Intelligent Surfaces (RIS) to effectively manipulate signal reflection or absorption to ensure system security. The proposed approach integrates feature selection, model selection, and an ensemble sequential model consultation strategy to deliver timely and robust detection of anomalous adversarial signals, along with corresponding RIS-based mitigation measures. Experimental results demonstrate that the proposed technique achieves up to 98.66% accuracy in detecting anomalies when evaluated against both existing benchmarks and simulated scenarios.
Paper
Full text
Machine Learning Intervened RIS-Based RF Interference Management for IoT
Semantic Scholar · Engineering · 2024
Abstract
The rapid proliferation of the Internet of Things (IoT) has garnered significant attention and is poised to profoundly transform daily life. IoT devices offer a wide range of applications, spanning from smart homes to sophisticated business networks and medical care systems. However, this widespread adoption also exposes IoT devices to potential security threats and malicious attacks, posing a persistent risk to the safety of sensitive data they handle. The consequences of security breaches extend beyond individual victims to affect the broader global community. In response to this challenge, machine learning (ML) techniques have emerged as promising tools for enhancing IoT device security. This paper proposes a novel unsupervised ML-based framework designed to efficiently detect anomalous signals, and thereby enabling active Reconfigurable Intelligent Surfaces (RIS) to effectively manipulate signal reflection or absorption to ensure system security. The proposed approach integrates feature selection, model selection, and an ensemble sequential model consultation strategy to deliver timely and robust detection of anomalous adversarial signals, along with corresponding RIS-based mitigation measures. Experimental results demonstrate that the proposed technique achieves up to 98.66% accuracy in detecting anomalies when evaluated against both existing benchmarks and simulated scenarios.