Code for the publication "Low-Frequency Black-Box Backdoor Attack via Evolutionary Algorithm"

This research aims to investigate the vulnerabilities of existing convolutional neural networks (CNNs) and vision transformers (ViTs) against backdoor attacks and to develop a novel backdooring approach. The study focuses on advancing a new technology in this area. The research uses textual data, with all data (i.e., source code) being independently developed.本研究旨在探究现有卷积神经网络（Convolutional Neural Networks，CNNs）与视觉Transformer（Vision Transformers，ViTs）在后门攻击（Backdoor Attacks）下的脆弱性，并开发一种新型后门植入方法（Backdooring Approach）。本研究聚焦于推动该领域的技术发展。研究采用文本数据，所有相关数据（即源代码）均为自主研发。