一种生成对抗样本、训练流量检测模型的方法及装置
By denoising the original traffic and generating adversarial perturbations using shadow models, the problem of noise affecting the quality of adversarial samples in Tor traffic is solved, thus improving the robustness and security of the detection model.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING TOPSEC NETWORK SECURITY TECH
- Filing Date
- 2023-09-22
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies fail to effectively handle network noise in Tor traffic when generating adversarial examples, resulting in low quality adversarial examples that affect the robustness of detection models. Furthermore, existing methods are prone to exposing model structure and parameters, posing a risk of being attacked by hackers.
By denoising the original traffic, using the shadow model to obtain gradient information, generating adversarial perturbations, and combining feature weights and shadow images to calculate adversarial examples, we can avoid directly attacking the detection model and protect the model's information security.
It improves the quality and transferability of adversarial examples, enhances the robustness of the detection model, reduces the risk of attacks, and improves the security of the model.
Smart Images
Figure CN117114058B_ABST