Security: Intrusion Prevention System Using Deep Learning on the Internet of Vehicles Vipparthy Praneeth * , Kontham Raja Kumar, Nagarjuna Karyemsetty Dept. of Computer Science and Systems Engineering, Andhra University College of Engineering (A), Andhra University, Visakhapatnam, India Corresponding Author Email: dr.krkumar@andhrauniversity.edu.in https://doi.org/10.18280/ijsse.110303 ABSTRACT Received: 9 April 2021 Accepted: 10 June 2021 Internet of vehicles supports to transfer of safety-related messages, which help to mitigate road accidents. Internet of vehicles allows vehicle to cooperative communicate, share position and speed data among vehicles and road side units. The vehicular network become prone to large number of attacks including false warnings, mispositioning of vehicles etc. The authentication of messages to identify the normal message packet from attack messages packet and its prevention is a major challenging task. This paper focuses on applying deep learning approach using binary classification to classify the normal packets from malicious packets. The process starts with preparing the training dataset from the open-source KDD99 and CICIDS 2018 datasets, consisting of 1,20,223 network packets with 41 features. The one-dimensional network data is preprocessed using an autoencoder to eliminate the unwanted data in the initial stage. The valuable features are then filtered as 23 out of 41, and the model is trained with structured deep neural networks, then combined with the Softmax classifier and Relu activation functions. The proposed Intrusion prevention model is trained and tested with google Colab, an open platform cloud service, and the open-source tensor flow. The proposed prevention classifier model was validated with the simulation dataset generated in network simulation. The experimental results show 99.57% accuracy, which is the highest among existing RNN and CNN-based models. In the future, the model can be trained on different datasets, which will further improve the model's efficiency and accuracy. Keywords: a deep neural network, Google CoLab, Internet of Vehicles, network intrusion prevention, security, vehicular network, CICIDS2018 dataset 1. INTRODUCTION The Internet of Vehicles (IoV) plays a vital role in communicating the safety-related message, safeguarding drivers, passengers, people, and the vehicle itself. Other than traditional wired networks have security mechanisms such as gateways, firewalls, and so on, wireless vehicular networks are still vulnerable to safety attacks that threaten nearly the entire infrastructure. VANETs (Vehicular Ad-hoc Network) are sub- division of IoV, operates based on ad-hoc mode, and can be exposed to various sensitive and unsafety actions, including manipulation of communications, send-away, spamming, masquerading, etc. Implementation of security [1] in IoV was found to be one of the significant challenges. To do so correctly and efficaciously, it must comply with the security specifications for protection against attackers and malicious vehicle nodes. To detect intruders is one crucial step in ensuring the security of vehicular networks. Several intrusion detections and prevention methods are available based on statistical analysis, cluster analysis, artificial neural networks, or deep learning. Due to self-learning and the adaptive nature of deep learning, it is a preferable method in intrusion detection and prevention. Malicious behavior in vehicles network can be detected by connecting and interacting with cameras (CAM), ad hoc vehicle networks, and roadside supporting equipment, and the vehicular node itself. An Intrusion Detection System (IDS) [2] may be considered a safer approach to identifying intruders. The IDS demands that each packet obtained or transferred between vehicle nodes is collected and thoroughly examined. Using test data from the IoV, the protection system can detect normal and even malicious behavior. Figure 1 illustrates the typical Internet of vehicle architecture [3], which contains vehicles having On-Board Unit (OBU) software integrated with the Intrusion Detection System. In this setup, the vehicular network consists of three normal vehicles with integrated IDS in OBU1 to OBU3 and one vehicle in intruder vehicle OBU4 (represented in red). An open-source Network simulator (NS2 2.34) [4] tool was used to simulate from low to extensive networks, as shown in Figure 2. Simulation of Urban MObility (SUMO) [5] open- source simulator used to generate road structures and network traffic. Various simulation scenario such as low, medium and high density networks were simulated with varying network parameters including size of networks, routing protocols, communication range, packet size etc. Table 1 shows the output file generated in network simulation which comprise of type of packet, transmission time, packet size, protocol, source and destination address and message etc. Deep Learning (DL) [6] is a subset of the machine learning (ML) branch that, in effect, shapes the AI branch. DL consists of several hidden neural network layers, one input, and one output layer. This increases efficiency by growing the data collection, i.e., learning a lot more than machine learning with more datasets. Deep learning also automatically considers all miniature features of the Dataset and chooses the most relevant and continuous learning. This also uses different neural International Journal of Safety and Security Engineering Vol. 11, No. 3, June, 2021, pp. 231-237 Journal homepage: http://iieta.org/journals/ijsse 231