An adaptive trust threshold based on Q-Learning for detecting intelligent attacks in vehicular Ad-Hoc Networks

Due to the intelligence of attack, some malicious nodes of the Vehicular Ad-Hoc Networks (VANETs) can evade detection and reconnaissance, which poses a huge security threat to the network security. With considering the sufficient adaptability and limited resources consumption in a small state space, a Q-Learning based adaptive trust threshold control strategy (QART) is proposed to balance the detection efficiency of the malicious vehicle and the false alarm of the normal vehicle. Compared with the existing intelligent attack detection schemes, the detection efficiency of the malicious vehicle is higher and the false alarm of the normal vehicle is lower under the proposed strategy. Finally, the experimental results verify that the proposed strategy can identify malicious vehicles in time and effectively reduces false alarms.