Online Criminal Behavior Recognition Based on CNNH and MCNN-LSTM

Abstract

In light of the proliferation of cybercrimes, the effective identification and mitigation of such online criminal activities has emerged as a significant challenge within the domain of network security. Therefore, this study introduces dilated convolution technology, self-attention mechanism, convolutional neural network and long short-term memory network, and proposes an overlapping traffic recognition model based on improved convolutional neural network and an online crime recognition model with long short-term memory network. In the traffic segmentation model test, the recall rate, F1 value, and error rate of the model under normal traffic conditions were 91.43%, 93.46%, and 92.43%, respectively. The error rate was 4.15%. The accuracy of the online crime recognition model for malware propagation and illegal transactions was 96.54% and 92.87% respectively. In the concept drift test, when the training time and test time interval was 60 days, the accuracy of the model was 48.67% higher than that of the long short-term memory network. Compared with the mainstream framework and traditional methods, its accuracy in high traffic scenarios was 94.78%, the error rate was 3.89%, and the P-value was < 0.05. In the final simulation test, the model could effectively identify illegal software transactions. The results show that the proposed model has high accuracy and strong generalization ability in identifying overlapping traffic and website fingerprint crimes, and effectively improves the detection ability of criminal activities in anonymous networks.