AI-Powered Intrusion Detection Systems for Securing Smart Home Networks. A Machine Learning

The widespread use of smart devices in our homes has led to highly interconnected systems that are increasingly vulnerable to cyberattacks. Conventional security approaches, originally built for static enterprise environments, are inadequate in addressing the demands of dynamic and resource-limited smart home networks. This study responds to the need for adaptable, lightweight, intelligent security measures by introducing an AI-driven Intrusion Detection System (IDS) designed for smart home environments. The work aims to design and assess a machine learning-based IDS capable of accurately identifying three common types of attacks: Brute Force login attempts, denial-of-service (DoS) attacks, and Device Hijacking. The system is trained and validated using publicly available datasets: CIC-IDS2017 and IoT-23. The development process follows a methodology that includes threat modelling, data preprocessing, feature selection, model training using Random Forest and Long Short-Term Memory (LSTM) networks, performance evaluation, and result interpretation. Furthermore, the system is developed with edge deployment in mind, ensuring compatibility with devices such as Raspberry Pi used in smart home gateways. Key contributions of the research include: (1) a targeted IDS framework tailored for smart home environments, (2) a performance comparison between temporal and ensemble learning models, and (3) a strategic plan for scalable and adaptive edge implementation. Results indicate strong detection accuracy and promising suitability for real-time use in Internet of Things (IoT) home environments. This study enhances the field of AI-based cybersecurity by offering an approach to safeguarding smart homes from emerging cyber threats, with meaningful implications for privacy, personal safety, and national cyber defence efforts.