Adversarial sample generation method and device, electronic equipment and storage medium

Abstract

Embodiments of the invention provide an adversarial sample generation method and device, electronic equipment and a storage medium. The method comprises the steps of obtaining an original text; determining a replacement word candidate set of each word in the original text; and based on a particle swarm optimization algorithm, searching a sample of the attack target model from a discrete space formed by the combination of the replacement word candidate sets, and generating an adversarial sample. According to the embodiment of the invention, the particle swarm optimization algorithm is used forsearching the adversarial sample, and the particle swarm optimization is more efficient than the genetic algorithm as a meta-heuristic group evolution calculation method, so that the search speed canbe increased when the algorithm is used for searching the adversarial sample, and the attack success rate can also be increased. For different natural language processing models, the embodiment of theinvention can quickly and efficiently generate a large number of high-quality confrontation samples, successfully cheat the target model and further expose the vulnerability of the target model, andhas good practicability.

Description

technical field [0001] The present invention relates to the technical field of natural speech processing, in particular to a method, device, electronic equipment and storage medium for generating an adversarial example. Background technique [0002] Adversarial attack refers to the process of making the judgment of the target model wrong by generating adversarial samples. Adversarial attacks can expose the vulnerability of machine learning models, thereby improving the robustness and interpretability of the models. Text adversarial attack refers to the process of generating adversarial samples by modifying the original text to make the judgment of the natural language processing model wrong. [0003] Existing research shows that deep learning models are highly vulnerable to adversarial attacks, such as simple modifications to abusive texts that can fool state-of-the-art abusive detection systems. In view of the fact that the natural language processing model based on deep ...

AI Technical Summary

Problems solved by technology

[0005] The embodiment of the present invention provides a method, device, electronic device and storage medium for generating an adversarial example, to solve the problem of the low speed of the search algorithm in the prior art and the The problem of low attack success rate

Images