Biblio

Securing Internet of Things (IoT) systems is a challenge because of its multiple points of vulnerability. A spate of recent hacks and security breaches has unveiled glaring vulnerabilities in the IoT. Due to the computational and memory requirement constraints associated with anomaly detection algorithms in core networks, commercial in-line (part of the direct line of communication) Anomaly Detection Systems (ADSs) rely on sampling-based anomaly detection approaches to achieve line rates and truly-inline anomaly detection accuracy in real-time. However, packet sampling is inherently a lossy process which might provide an incomplete and biased approximation of the underlying traffic patterns. Moreover, commercial routers uses proprietary software making them closed to be manipulated from the outside. As a result, detecting malicious packets on the given network path is one of the most challenging problems in the field of network security. We argue that the advent of Software Defined Networking (SDN) provides a unique opportunity to effectively detect and mitigate DDoS attacks. Unlike sampling-based approaches for anomaly detection and limitation of proprietary software at routers, we use the SDN infrastructure to relax the sampling-based ADS constraints and collect traffic flow statistics which are maintained at each SDN-enabled switch to achieve high detection accuracy. In order to implement our idea, we discuss how to mitigate DDoS attacks using the features of SDN infrastructure.

Securing Internet of Things is a challenge because of its multiple points of vulnerability. In particular, Distributed Denial of Service (DDoS) attacks on IoT devices pose a major security challenge to be addressed. In this paper, we propose a DNS query-based DDoS attack mitigation system using Software-Defined Networking (SDN) to block the network traffic for DDoS attacks. With some features provided by SDN, we can analyze traffic patterns and filter suspicious network flows out. To show the feasibility of the proposed system, we particularly implemented a prototype with Dirichlet process mixture model to distinguish benign traffic from malicious traffic and conducted experiments with the dataset collected from real network traces. We demonstrate the effectiveness of the proposed method by both simulations and experiment data obtained from the real network traffic traces.