Network security traffic detection and legal supervision based on adaptive metric learning algorithm

• This paper proposes an adaptive algorithm of metric learning based on deep learning • the vertex weighting algorithm proposed in this paper has better robustness and generalization • It is necessary to break away from the original local limitations and take a comprehensive view. Network security mainly utilizes relevant equipment and programs of network systems to protect private information, in order to prevent damage, tampering, or leakage by illegal elements, and ensure the smooth operation of network system security. This paper proposes an adaptive algorithm of metric learning based on deep learning, which classifies and learns the source domain images, and adds and modifies the classification margin to enrich the aligned classification boundary images. After a series of experimental analysis, it was found that the vertex weighting algorithm proposed in this paper has better robustness and generalization compared to other algorithms, achieving better classification of target area images. Using the vertex weighted function algorithm, we can analyze the Time complexity of the original algorithm and calculate more accurate time values. Criminal law has a certain degree of lag, and its manifestation in cybersecurity crimes is relatively limited. It is necessary to break away from the original local limitations and take a comprehensive view. Overall, traditional criminal law theories and principles have high practical value and significance.