ML-BASED MITIGATION OF UDP FLOODING ATTACKS IN LOT

Abstract

Internet of Things (IoT) is the concept that is directing a huge increase in the Internet and its capability to collect, investigate and distribute data which can be turned into information or knowledge. One category of devices such as the Low Power Lossy networks (LLNs) consists of numerous small sensors and low power devices as building block elements in IoT. The ROLL working group at Internet Engineering Task Force (IEFT) has designed the Routing Protocol for LLNs (RPL) which is the core of the IoT protocol stack used for communication between these low-power devices. IoT devices are resource constrained devices and hence it is very easy to exhaust them of their resources or deny availability. One of the most prominent attacks on the availability is the Denial of service (DoS) attack. Although, DoS is not a new Internet attack but in the recent times of 2017-18 even bigger DoS attacks took place. A few simulation tools exist which enable evaluation of RPL for a realistic deployment scenario. This paper focuses on understanding of the implementation of UDP flood attacks on RPL then recognizing the attacker nodes using statistical based approach for detecting Local Outlier. The implementation and analysis is carried out by the simulations done in the Contiki OS Cooja simulator with respect to the performance metrics such as radio duty cycle, energy consumption. Furthermore, in this paper, an ML based approach is proposed that could be used recognize the attacker to mitigate UDP flood attacks.