An anti-reconnaissance evasion attack method for deep person re-identification system

Abstract

The invention discloses an anti-reconnaissance evasion attack method oriented to a deep pedestrian re-identification system, and proposes an optimization method based on the maximization of matching differences, and combines multi-position sampling to generate noise patterns that are variable across cameras and can be expanded in position, so that pedestrians Any position in the monitoring area of ​​the re-identification system, the same noise pattern cannot be matched with each other under different camera shots. In addition, this method integrates physical environment factors into the noise pattern generation process, reduces noise information loss during printing and shooting, and improves its robustness. The noise pattern generated by this method can prevent the pedestrian re-identification system from correctly searching and locating the attacker, and realize "invisibility" under the security monitoring system.

Description

technical field [0001] The present invention designs the field of artificial intelligence security, and specifically refers to an anti-reconnaissance and escape attack method oriented to a deep pedestrian re-identification system. Background technique [0002] With the rapid development of the mobile Internet, the continuous upgrading of hardware equipment, the production of massive data and the updating of algorithms, the development of artificial intelligence (AI) has become unstoppable, and it is gradually penetrating and profoundly changing human life. At present, artificial intelligence technology based on machine learning and deep learning is widely used in various fields such as human-computer interaction, visual processing system, recommendation system, safety diagnosis and protection, and its application scenarios include unmanned driving, image recognition, malware detection, Malware filtering and more. It can be said that the advent of the era of artificial intel...

AI Technical Summary

Problems solved by technology

Once the deep person re-identification system is vulnerable to specific attacks, it will bring serious consequences and security threats, for example, criminals can evade the search and location of law enforcement agencies, or spies can infiltrate classified classified areas under surveillance

Images