Resumen
The recent surge in Internet of Things (IoT) deployment has increased the pace of integration and extended the reach of the Internet from computers, tablets and phones to a myriad of devices in our physical world. Driven by the IoT, with each passing day, the Internet becomes more integrated with everyday life. While IoT devices provide endless new capabilities and make life more convenient, they also vastly increase the opportunity for nefarious individuals, criminal organizations and even state actors to spy on, and interfere with, unsuspecting users of IoT systems. As this looming crisis continues to grow, calls for data science approaches to address these problems have increased, and current research shows that predictive models trained with machine learning algorithms hold great potential to mitigate some of these issues. In this paper, we first carry out an analytics approach to review security risks associated with IoT systems, and then propose a machine learning-based solution to characterize and detect IoT attacks. We use a real-world IoT system with secured gate access as a platform, and introduce the IoT system in detail, including features to capture security threats/attacks to the system. By using data collected from a nine month period as our testbed, we evaluate the efficacy of predictive models trained by means of machine learning, and propose design principles and a loose framework for implementing secure IoT systems.