Malicious adversarial sample generation method and device based on evolutionary generative adversarial network

The invention discloses a malicious adversarial sample generation method and device based on an evolutionary generative adversarial network, and the method comprises the steps: generating a malicious software adversarial sample through a generator in the evolutionary generative adversarial network, and combining the malicious software adversarial sample with a binary malicious sample of real malicious software; cuckoo sandbox is used to test the performability and functionality of the generated malicious software confrontation samples, a malicious sample data set is constructed, a discriminator is used to detect and identify, a loss function is transmitted to a generator, parameter updating and training of the generator are guided, the malicious confrontation samples are generated and then screened by using a genetic algorithm, multiple iterations are carried out, and the malicious software confrontation samples are obtained. Therefore, malicious software confrontation samples with excellent performance are g