Game Theoretic Study On Security In Vehicular Networks

Vehicular network is a kind of wireless networks with high mobility and is large in scale with sensors,drivers,mobile terminals,base stations and servers,which is vulnerable to adversaries such as faked attacks,jamming attacks,spoofing attacks and so on.Drivers carry out faked sensing attacks defrauding server of its payment,which severely decreases quality of the mobile crowdsensing.Jammers send interfering signal to jam legal communication in vehicular networks,aiming to paralyse the vehicular network services.Spoofers send spoofing packets with faked identities for illegal advantages and performing further attacks,such as denial-of-service attacks.Therefore,the vehicular network security is becoming an important factor that limits its growth in the future.Firstly,the interactions between a crowdsensing server and vehicles are formulated as a game to reduce the probability of carring out faked sensing attacks.Nash equilibrium of the game is derived for both accumulative sensing tasks and best-quality sensing tasks,revealing the relationships between the crowdsensing quality and wireless environment.A reinforcement learning based payment strategy is proposed to reduce the attack rate and improve quality of the sensing reports.For example,compared with the greedy payment strategy,the proposed payment strategy in best-quality sensing tasks with 80 vehicles improves quality of the sensing reports and server's utility by 72%and 233%,respectively.Secondly,the power control strategy of a vehicle against a jammer is formulated as an anti-jamming game.The vehicle determines its transmit power in advance,and the jammer chooses the jamming power according to the observed ongoing transmission.Stackelberg equilibrium and Nash equilibrium of the anti-jamming game are derived,and the anti-jamming performance of the vehicle is investigated with these two equlibria.Simulation results show that the jammer with Stackelberg equilibrium has stronger ability of attacking.However,with the observation error of the jammer regarding the transmit power of the vehicle increasing,the SINR and utility of vehicle increase by 12%and 20%,respectively.Finally,the interactions between a server and a spoofer are formulated as a physical(PHY)-layer authentication game.The spoofer chooses its attack rate,and the server chooses the test threshold.NE of the PHY-layer authentication game is derived,and the condition that the NE exists is presented to reveal the relationships between wireless environment and spoofing detection accuracy.The PHY-layer spoofing detection based on Q-learning is proposed for dynamic environment,and Dyna-PS algorithm is further applied to accelerate learning speed and improve the detection accuracy.For example,compared with Q-learning algorithm,Dyna-PS based PHY-layer authentication scheme decreases the miss detection rate and false alarm rate by 31.5%and 5.6%,respectively.