Pushing intelligence to the network edge
Ola Salman, Louma Chaddad, Imad H. Elhajj, Ali Chehab, Ayman Kayssi · 2018
Networking has already entered a new era where it is evolving into the Internet of Things (IoT). The future network consists of tens of billions of connected devices. These things bear different security vulnerabilities and can be abused to perform diverse attacks. Due to the heterogeneous nature of IoT and its massive deployment, managing the IoT security requires automated techniques to deal with the emerging security challenges. To address this problem, we propose a model where intelligence is integrated at the network edge to extract specific statistical features per network flow. These features are passed to a controller which implements a classification algorithm to determine the device type generating the observed flow. A monitoring module is responsible for analyzing the network traffic behavior. If abnormal traffic is detected from a certain device, the controller takes the adequate action (e.g. blocking the traffic from this device). After collecting data from IoT devices, we evaluate our proposed classification model. Results show that our approach can achieve up to 99.6% accuracy for device type identification and up to 99.9% for traffic type classification.