Voiceprint recognition attack defense method based on cuckoo search algorithm

Abstract

The invention discloses a voiceprint recognition attack defense method based on a cuckoo search algorithm. The method comprises the following steps of (1) preparing a raw audio data set; (2) traininga voiceprint recognition model, wherein the voiceprint recognition model is trained by using a pre-training data set, and then a test data set is used to test the accuracy of the recognition model; (3) attacking the voiceprint recognition model, wherein an attack method based on the cuckoo search algorithm is built, a fitness function and related parameters of the attack method are set, and an optimal adversarial sample is generated by using the attack method and is mistakenly identified as a target category to be recognized by the human ears; and (4) performing adversarial training of the voiceprint recognition model, wherein the sample generated in the step (3) is added into the pre-training data set, and the voiceprint recognition model is re-trained, so that the re-trained voiceprint recognition model has the capability of defending attack of the adversarial sample, and the safety and stability of the voiceprint recognition model are improved.

Description

Technical field [0001] The invention belongs to the field of deep learning algorithms and information security research, and specifically relates to a voiceprint recognition system attack defense method based on a cuckoo algorithm. Background technique [0002] Voiceprint recognition, as a biometric technology, can effectively distinguish the voices of different people by extracting unique features in the speaker's voice. Because the voice containing voiceprint features is easy to obtain, and the recognition algorithm is relatively low in complexity, the voiceprint recognition system has a wide range of applications. The existing voiceprint recognition technology is divided into text-related and text-independent. The text-related voiceprint recognition model requires users to pronounce according to the specified content. Each person's voiceprint model is accurately established, but the recognition must also follow the specified content. Pronunciation can achieve a better recogni...

AI Technical Summary

Problems solved by technology

[0005] The present invention overcomes the security problem that the existing voiceprint recognition system is vulnerable to attack by adversarial samples, and provides a voiceprint recognition attack defense method based on the cuckoo search algorithm, which can efficiently generate adversarial samples to make the recognition system misidentify, And use the way of confrontational training to add the generated confrontational samples into the original training data set. The voiceprint recognition model after confrontational training can defend against the attack of confrontational samples and has higher security.

Images