Smart Home Networks: Security Perspective and ML-based DDoS Detection

• Published in: Conference proceedings - Canadian Conference on Electrical and Computer Engineering pp. 1 - 8
• Main Authors: Al Mtawa, Yaser, Singh, Harsimranjit, Haque, Anwar, Refaey, Ahmed
• Format: Conference Proceeding
• Language: English
• Published: IEEE 30.08.2020
• Subjects: Computer crime; home security threat; IEEE 802.15 Standard; IoT; ML-based algorithms; network security; Operating systems; Protocols; Radiofrequency identification; Security; security system; smart home; Smart homes
• ISSN: 2576-7046
• DOI: 10.1109/CCECE47787.2020.9255756

Internet of Things (IoT) allows households to have real-time access to various services such as IP cameras for monitoring home security, automatic climate controls, doors and window access for family members including pets, mood-based home lighting, or automatic refrigerators that tell what item is going to expire at what time, etc. However, IoT has often been considered a single-domain facility that provides services to consumers, but security and privacy issues pertaining to households and residential infrastructure have not been studied thoroughly. The security of IoT -based systems requires a critical engineering infrastructure when dealing with home security. Any proposed home security framework should establish an automatic access monitoring and regular updates of the system software and firmware according to the ongoing threats. In this paper, we provide an insightful short review of security issues in smart home systems. We highlight several popular IoT protocols and provide critical analysis of their performance against various security threats. Mitigation strategies are also presented to reduce the impact of cyberattacks on smart home systems. Furthermore, our analysis shows that DDoS is one of the most common threats to smart home systems. Therefore, we developed an ML-based model to detect DDoS attacks. We employed unique characteristics of IoT traffic to engineer features that enable ML algorithms to accurately classify DDoS traffic from normal/benign traffic.