Misbehaviour Detection based on Least Square Twin Support Vector Machine in VANETs

A great deal of study has not been done on machine learning techniques to identify inappropriate activities in Vehicular Adhoc Networks (VANETs). Unauthorized vehicles have the likely to send erroneous messages across VANET, which can lead to a difficult situation that cannot be avoided. This article describes a machine learning technique to classify misbehaviours in VANET. The proposed strategy is based on the Least Square Twin Support Vector Machine (LSTSVM), which the authors developed. Two of the mechanisms developed are called data based trust models (DTM) and vehicle based trust models (VTM). The data based trust model is responsible for identifying fraudulent messages by employing a LSTSVM-based classifier that evaluates message content in conjunction with vehicle characteristics. The message propagation behaviour of the vehicle is evaluated by the LSTSVM-based classifier utilized by the local VTM, which evaluation determines whether or not the vehicle may be trusted. Then, using the Dempster–Shafer theory, the trust authority (TA) VTM will compile several different trust assessment reports that pertain to the same vehicle and provide a comprehensive trust value. Compared to base classifiers utilized for classification, the LSTSVM approach is more effective in classifying a wide variety of inappropriate behaviours that can be discovered in VANET.