Research on Feature Extraction Strategies for Cybercrime Crimes Combined with Deep Learning and Their Probabilistic Models

In this paper, in order to improve the accuracy and precision of the detection and identification of cybercrime cases, eliminating the incomplete and ambiguous information obtained from a single source of evidence, combined with the procedural steps of the sampling and forensics for the detection of cybercrime cases, a cybercrime detection and control model based on the fuzzy reasoning and the improvement of the D-S algorithm is proposed. Combined with the application of the improved D-S evidence theory in the fusion of cyber evidence, the weighted Bayes is combined to achieve effective forensics of cybercrime evidence. The improved D-S evidence theory algorithm achieves a detection rate of 0.989 for network anomalous data, which has an obvious advantage in network data fusion over the no-preprocessing, original D-S evidence fusion algorithm. The intrusion path generation algorithm considers the intrusion risk of intrusion path IV5 to be the highest at 0.186 with the intrusion paths <H1, W2, S2, S3, W5, S3, T5>.