Mitigating DoS attack in MANETs considering node reputation with AI

Abstract

Mobile Adhoc Network (MANET) is a decentralized and dynamically adoptable network. It is infrastructure less and hence can be used where a fixed configuration is not possible or required. MANETs have various real-life applications and hence have gained the attention of research community. Security is an integral part of any computer network system and MANETs are no different. This paper focuses on solving DoS attacks in MANET and shows that a general classification model might fail to identify this kind of attacks as these models fail to differentiate between network errors and a real DoS attack. A reputation-based node classification scheme is proposed to improve identification of real DoS attacks versus any other cause that might not be an attack. Results showed that our proposed reputation-based approach when integrated with any classifier increases its accuracy by around 3.25%. Further, the combined model is able to block real DoS attacks and allow any other cause which is not an attack.