Mobile Edge Computing Task Offloading Strategy And Security Performance Research Based On Online Learning Algorithm

With the popularity of mobile devices and the rapid growth of mobile applications,Mobile edge computing(MEC)has become a new computing model.MEC transfers computationally intensive tasks to nearby servers for task offloading,alleviating computing pressure and reducing mobile users’ energy consumption and latency.This technology is widely used in mobile application scenarios such as smart home and the Internet of Vehicles.Compared to traditional cloud computing models,MEC has the advantages of low latency,high reliability,and high bandwidth,solving the problems of high latency and network congestion between cloud computing network data centers and users.In MEC systems,wireless devices(WDs)have different real-time computing tasks,and each task needs to be processed locally or offloaded to an edge server.Different task offloading schemes have a significant impact on task completion latency and mobile device energy consumption.It is necessary to develop resource management and scheduling strategies that adapt to dynamic environmental changes to improve the user’s sense of experience;on the other hand,due to the broadcast nature of wireless channels,computing tasks offloaded by terminal devices to edge servers may be stolen by malicious attackers,resulting in the loss or disclosure of user privacy.Therefore,designing a reasonable task offloading strategy and ensuring data transmission security is particularly important.To solve the above problems,this paper focuses on mobile edge computing,and studies the methods of task unloading and server allocation under the premise of ensuring efficient and secure transmission of tasks.The specific research content is as follows:1)In order to solve the problem of computing offloading and scheduling multiple edge servers in MEC,this paper constructs a convolutional neural network framework,which integrates priority,data size,bandwidth,and channel state information factors,and proposes a distributed unsupervised learning algorithm(DULO).The performance of the algorithm is tested in single server and multi server scenarios for multi-user and multi task scenarios.The experimental results show that compared to other algorithms,the algorithm proposed in this paper outperforms other offloading schemes in terms of user energy consumption and task offload delay performance,and can generate an asymptotically optimal offload decision in a timely manner.2)To address the security issue of task offloading that is prone to eavesdropping,this paper proposes an online offloading algorithm based on physical layer security technology.Without considering network prior knowledge,a traversal secret queue is constructed to minimize the average energy consumption of devices.Using Lyapunov optimization method,the CPU cycle frequency,transmission power,and task offloading issues are decomposed into multiple simple subproblems.Simulation results confirm that this method is significantly superior to the benchmark method in terms of throughput and energy consumption.By increasing the gap between the average transmission rate and the eavesdropping rate,the security of the system can be improved.