___________________________________________________________________________________________________ AbstractThe phenomenal and continuous growth of diversified IOT (Internet of Things) dependent networks has open for security and connectivity challenges. This is due to the nature of IOT devices, loosely coupled behavior of internetworking, and heterogenic structure of the networks. These factors are highly vulnerable to traffic flow based DDOS (distributed-denial of services) attacks. The botnets such as "mirae" noticed in recent past exploits the IoT devises and tune them to flood the traffic flow such that the target network exhaust to response to benevolent requests. Hence the contribution of this manuscript proposed a novel learning-based model that learns from the traffic flow features defined to distinguish the DDOS attack prone traffic flows and benevolent traffic flows. The performance analysis was done empirically by using the synthesized traffic flows that are high in volume and source of attacks. The values obtained for statistical metrics are evincing the significance and robustness of the proposed model

No abstract

Summary The term that makes physical objects and devices connect over each other through the wireless networks is called the Internet of Things (IoT). The IoT networks are highly vulnerable to DDoS attacks due to their property of dynamic adapting new devices, even though several existing contributions depicted DDoS defense techniques. However, the features deliberated for notifying DDoS attacks are not competent and constant to attain minimum false alarming and optimum detection accuracy. This article portrayed a novel method called regression coefficient of traffic flow metric (RCTFM) for DDoS defense in IoT networks. This method is used for performing predictive analysis in detecting the scope of DDoS attacks from transactions buffered during a specified static time frame. Unlike the contemporary models, the proposed model considers the traffic that buffered under a given time frame as input and derives regression coefficients of the parameters portrayed as metrics of the corresponding buffered traffic flow. The simulation study evinces the scalability and significance of the proposal that scaled against other existing methods.

No abstract