基于LKAN模型的恶意加密流量检测OA

With the rapid development of network technology,malicious encrypted traffic has become an important threat in the field of network security.Malicious encrypted traffic encapsulates malicious data through encryption technology,making it difficult to be identified and intercepted by traditional detection methods.This paper proposes a malicious encrypted traffic detection model named LKAN based on Long Short-Term Memory network(LSTM)and Kolmogorov Arnold Networks(KAN).LSTM can effectively capture the temporal features of traffic data,and KAN is a neural network based on the function decomposition theory,which can efficiently learn the complex structure of high-dimensional data.By combining the advantages of LSTM and KAN,LKAN conducts feature extraction and classification to achieve accurate identification of malicious encrypted traffic.The proposed LKAN model was used to conduct multi-classification experiments on the ISCX-VPN-NonVPN-2016 dataset,achieving an accuracy of 0.982 591,which demonstrates the model's effectiveness and provides a novel approach for designing malicious encrypted traffic detection methods.