Edge computing compensates for the traditional cloud computing data transmission overhead.However,limited storage and computing resources in the edge network restrict its ability to deploy complex security algorithms,making it more vulnerable to Distributed Denial of Service(DDoS)attacks.This paper proposes a task classification-based Attention-1D-CNN detection model,TCA1C,aimed at solving problems in current edge networks,such as low performance in detecting DDoS,lack of a task classification mechanism,and weak ability to deal with multi-attribute traffic.First,the model classifies the traffic in the communication link according to different offloading tasks such that the overall offloading security is not affected when some tasks are attacked,and the attribute values of the traffic under the same task are extracted and normalized.Next,the model inputs the processed data into an Attention-1D-CNN.Channel and spatial attention determine the contribution of data features for classification and eliminate redundant information below the feature threshold,thereby reducing the complexity of the model learning process and allowing the model to converge quickly.The simulation results show that the accuracy of the TCA1C model in DDoS detection is as high as 99.73%,and the performance of the TCA1C model is better than those of the DT,ELM,LSTM,and CNN,with reduced detection time.When different offloading tasks face certain attack probabilities,traffic classification can effectively reduce the mutual influence of different tasks such that the computing tasks of the terminal equipment can maintain a high level of security during offloading.
edge computingDistributed Denial of Service(DDoS)attack detectiontask classificationattention mechanism1D-CNN module
国家科技期刊平台
NSTL
万方数据
维普
