Research On Intelligence Privacy Protection Technology Based On Reinforcement Learning In Internet Of Things

The rapid development of the Internet of Things(IoT)has given birth to the emergence of service models such as intelligent transportation,telemedicine,and location-based services.However,they face different privacy issues,such as eavesdropping attack and advance persistent threat,which has caused huge economic losses and even harm national security.Additionally,these new service models face challenges such as the complexity of the IoT environment,the unknown privacy leakage model,the limited resources of the IoT devices,and the time-sensitive requirements.This paper investigates the privacy protection schemes in three typical scenarios of cloud storage systems,mobile edge computing,and location-based services for privacy leakage in the IoT.We apply reinforcement learning,game theory,and differential privacy to investigate intelligent privacy protection scheme.This work provides a theoretical basis and useful suggestions for the future design of the IoT privacy protection framework.The main contributions of the paper are as follows.1.To defend against the advanced persistent threat(APT)in cloud storage systems,the interactions between an APT attacker and a defender allocating their constraint resources over multiple storage devices in a cloud storage system are formulated as a Colonel Blotto game.The Nash equilibria of the APT defense game are derived to evaluate how the system parameter impacts the performance.An APT defense scheme based on policy hill-climbing is proposed for the defender to achieve the optimal APT defense performance in the dynamic game without being aware of the APT attack model and the data storage model.This scheme selects the defense policy based on a mixed strategy table,improving the APT defense performance by introducing the randomness.A deep Q-network-based APT defense scheme further improves the APT detection performance for the case with a large number of CPUs and storage devices.Simulation results show that,compared to the Q-learning-based APT defense scheme,our proposed deep reinforcement learning-based scheme can improve the data protection level by 30.5%.2.To solve the privacy leakage issues in the mobile edge computing system,we propose a reinforcement learning-based privacy-aware offloading scheme to help healthcare IoT devices protect both the user location privacy and the usage pattern privacy.This scheme enables a healthcare IoT device to choose the offloading rate that improves the computation performance,protects user privacy,and reduces the energy consumption without being aware of the privacy leakage,IoT energy consumption,and edge computation model.This scheme uses a Dyna architecture and a post-decision state learning method to accelerate the learning process.We provide the performance bound of this scheme regarding the privacy level,the energy consumption,and the computation latency for three typical healthcare IoT offloading scenarios.Simulation results show that this scheme improves the privacy level by 36.6%,reduces 9.6%of the energy consumption,and decreases 68.8%of the computation latency compared with the benchmark CMDP scheme.3.To protect the location privacy in location-based service,we propose a reinforcement learning-based sensitive semantic location privacy protection scheme.This scheme uses the idea of differential privacy to randomize the released vehicle locations and adaptively selects the perturbation policy based on the sensitivity of the semantic location and the attack history.This scheme enables a vehicle to optimize the perturbation policy in terms of the privacy and the quality of service(QoS)loss in a dynamic privacy protection process without being aware of the current inference attack model.To solve the location protection problem with high-dimensional and continuousvalued perturbation policy variables,a deep deterministic policy gradient-based semantic location perturbation scheme is developed.Simulations are carried out to demonstrate that the proposed scheme increases the privacy level by 30.8%and reduces the QoS loss by 15.6%,compared to the Q-learning-based semantic location protection scheme.