Machine Learning-based Intrusion Detection for IoT Devices in Smart Home

The Internet of Things (IoT) is increasingly providing people with objects to connect with the physical world, which plays an important role in people’s daily life. Although it has brought us great convenience, there are also suffered from security vulnerabilities and potential threats. Currently, the lack of protection mechanisms for IoT devices with limited resources makes it easy to be attacked. In this paper, we design an intrusion detection system to protect the IoT security. The system uses supervised learning to achieve two main functions: (1) classify the generated malicious traffic; (2) identify the types of attacks. Besides, we propose a lightweight feature selection method, which uses a small number of features to evaluate the two functions. As a result, in the classification experiments, the proposed method automatically extracts 29 and 9 from 88 features, and then the designed system achieves a high accuracy rate of 98.7% and 98.99%. This means that our method still has great accuracy by taking a small number of features.