Machine learning Approach for detection unknow activity in vehicular adhoc network

Abstract

An enabling technology for providing security and valuable information in modern transport systems but susceptible to a number of attacks, ranging from auditing passively to aggressive interference is Vehicular ad hoc network also termed as VANET. IDSs i.e. Intrusion Detection Systems are critical tools for risk reduction when suspicious activities are detected. Additionally, collaborations of VANET vehicles increases the accuracy in detection by communicating interactions among their nodes. Due to this, machine learning framework of distribution is effective and can be scaled and applied to build collaborative detection algorithms over VANETs. Concern about privacy is a fundamental obstacle to collaborative learning, because data is exchange between nodes. A node which is malicious can get information that is sensitive about nodes other than itself through the data which is observed. This transcript proposes for VANETs, a collaborative IDS that safeguards machine learning privacy. In the algorithm proposed, the alternating multiplier direction approach is used to a class of empirical risk minimization problems and an intrusion detection classifier is trained in the VANET. The use of privacy differential is done to capture the notation of privacy and apply a vector approach of dual disturbance to dynamically varying privacy.