Synthetic datasets generation for intrusion detection in VANET

Abstract

Inter-car network -- a VANET (vehicular adhoc network) -- enables wireless communication between vehicles (V2V) and vehicle-to-infrastructure (V2X). The main goal of VANET is to render safety and convenience on the road. VANET differs from traditional networks due to its unique characteristics such as a high speed of hosts movement, a quickly changing topology, a frequent installation and disconnection of communication links. For a lack of infrastructure and centralized management, it becomes vulnerable to misbehaviors that significantly threatens different aspects of the VANET security. VANET should provide adequate security measures for the protected cyberenvironment. One of the commonly known approaches to protect a network is an intrusion detection system (IDS) that inspects a behavior of traffic and network hosts looking for the signs of the security threats and generates the alarm for any detected security anomaly. To be effective, IDS has to be trained with an adequate dataset of samples of security threats, but such task-driven datasets have not been produced for VANET so far. This paper discusses our method of synthetic generating a dataset for VANET IDS. There is a generator that allows providing datasets applying a network simulator NS-3 when investigating various types of specific cyber attacks targeted at VANET. The paper presents the existing datasets, describes our method developed to solve the task, discusses the characteristics of the resulting dataset, and shows the outcomes of simulation. The synthetically generated datasets may be applied for training the machine learning-based VANET IDSs being used to detect security threats in new car-to-car adhoc networks.