Countermeasure defense method for countermeasure attacks based on artificial immune algorithm

Abstract

The invention discloses a confrontation defense method for confrontation attack based on an artificial immune algorithm. The method comprises the following steps: (1) obtaining an original picture sample set; based on the original picture sample, generating an adversarial sample based on an artificial immune algorithm; merging the original picture sample and the adversarial sample, and dividing into a training set and a test set; (2) training a picture classifier by using the training set and the test set to obtain a picture recognition model; and (3) identifying the to-be-identified picture by utilizing the picture identification model to realize confrontation and defense of picture identification. According to the invention, the technical problems of poor defense effect and low classifier identification accuracy in the prior art when attacks are resisted based on an artificial immune algorithm can be solved.

Description

technical field [0001] The invention relates to the field of confrontation defense, in particular to a confrontation defense method for confrontation attacks based on artificial immune algorithms. Background technique [0002] In recent years, deep neural networks have made great breakthroughs in many machine learning fields, such as image classification, object recognition, object detection, speech recognition, language translation, speech synthesis and other fields. However, despite the great success of deep neural networks in these fields, a large number of recent studies have proved that even the best deep neural networks can still be broken by adversarial examples. The discovery poses a serious threat to some critical security applications, such as autonomous vehicles, biometrics and surveillance systems. You can simply imagine that in an automatic car system, a picture that should have been recognized as a stop at a red light was attacked by an adversarial example and...

AI Technical Summary

Problems solved by technology

[0006] The present invention provides an adversarial defense method for adversarial attacks based on artificial immune algorithms to solve the technical problems of poor defense effect and low classifier recognition accuracy in the prior art when facing adversarial attacks based on artificial immune algorithms

Images