Poster: A Semi-Supervised Framework to Detect Botnets in IoT Devices.

The number of IoT devices is growing at a rapid pace and the misuse of the shared communication channels has led to a new security challenge caused by botnets. Botnets are compromised IoT devices that are not only able to attack other devices but are also able to spread the infection in the network. In this work, we propose a novel Neural Networks based framework that can detect botnets in IoT devices. The key features of our work include (1) data labelling with minimal supervision with very high accuracy; (2) dynamic network updates to allow learning new attacks not yet discovered; (3) low detection latency to detect such attacks in real-time; and (4) detecting zero-day attacks. The evaluation was done on a dataset containing nine IoT commercial devices infected with BASHLITE Mirai. The experimental results demonstrate the usefulness of the framework providing highly accurate results with low-latency.