Deep Learning Based Intrusion Prevention System in Vehicular Network

In the internet of vehicles, safety-based communication is carried out for prevention, mitigation, and alleviation accidents through cooperative messages, position sharing, exchange speed data between vehicle (nodes) corresponding roadside units. However, such networks are susceptible to false alarms mispositioning vehicles. It therefore imperative authenticate identify normal messages from aggressive incorrect messages. this context, paper has emphasized on a deep learning technique utilizing binary classification segregating malicious packets. The procedure initiated by preparation training datasets KDD99 CICIDS 2018-type open-source having 1,20,223 packets 41 features. An autoencoder used in preprocessing stage elimination undesirable right beginning. 23 salient features filtered 41. For models, structural neural network utilized along with Softmax classifier rectified linear unit (ReLU) activation functions. complete intrusion prevention (IP) mechanism further trained & tested Google co-labs as open platform cloud service tensor flow. Furthermore, simulation set developed network-simulating validation model. results experimentation have established an accuracy 99.57% greater than existing recurrent convolution Neural Network models. work future, various can be improve efficacy.