ABSTRACT
A DoS attack is the most severe attack on IoT and creates a crucial challenge for the detection and mitigation of such attacks. A DoS attack occurs at multiple layers of the IoT protocol stack and exploiting the protocol vulnerabilities disrupts communication. Traditional mechanisms employ single-layer detection of DoS attacks, which individually detect and mitigate attacks. However, it is essential to establish a general framework for detecting DoS attacks in a real-time environment and coping with diversified applications. This can be achieved by fetching attack features of multiple layers to create a pool of numerous attacks and then designing a system that detects the attack when fed with specific attack features. This chapter comprehensively analyzes the research gap in the DoS attack detection techniques proposed. Secondly, we offer a two-stage framework for DoS attack detection, comprising Fuzzy Rule Manager and Neural Network (NN), to detect cross-layer DoS attacks in real time. The Input Data Type (IDT) is derived using a fuzzy rule manager that can identify the type of input dataset as usual or attack in real time. This IDT is passed to the NN along with the real-time dataset to increase detection accuracy and decrease false alarms.
