基于黑盒测试框架的深度学习模型版权保护方法  

Copyright protection for deep learning models utilizing a black-box testing framework

作　　者：屈详颜 于静[1,2] 熊刚 盖珂珂[3] Qu Xiangyan;Yu Jing;Xiong Gang;Gai Keke(Institute of Information Engineering,Chinese Academy of Sciences,Beijing 100085,China;School of Cyber Security,University of Chinese Academy of Sciences,Beijing 100049,China;School of Cyberspace Science and Technology,Beijing Institute of Technology,Beijing 100081,China)

机构地区：[1]中国科学院信息工程研究所,北京100085 [2]中国科学院大学网络空间安全学院,北京100049 [3]北京理工大学网络空间安全学院,北京100081

出　　处：《网络安全与数据治理》2023年第12期1-6,13,共7页CYBER SECURITY AND DATA GOVERNANCE

基　　金：国家自然科学基金(62006222)。

摘　　要：当前生成式人工智能技术迅速发展,深度学习模型作为关键技术资产的版权保护变得越发重要。现有模型版权保护方法一般采用确定性测试样本生成算法,存在选择效率低和对抗攻击脆弱的问题。针对上述问题,提出了一种基于黑盒测试框架的深度学习模型版权保护方法。首先引入基于随机性算法的样本生成策略,有效提高了测试效率并降低了对抗攻击的风险。此外针对黑盒场景,引入了新的测试指标和算法,增强了黑盒防御的能力,确保每个指标具有足够的正交性。在实验验证方面,所提方法显示出了高效的版权判断准确性和可靠性,有效降低了高相关性指标的数量。With the rapid development of generative artificial intelligence technologies,the copyright protection of deep learning models has become increasingly important.Existing copyright protection methods generally adopt deterministic test sample genera-tion algorithms,which suffer from inefficiencies in selection and vulnerabilities to adversarial attacks.To address these issues,we propose a copyright protection method for deep learning models based on a black-box testing framework.This method introduces a sample generation strategy based on randomness algorithms,effectively improving test efficiency and reducing the risk of adversari-al attacks.Additionally,new test metrics and algorithms are introduced for black-box scenarios,enhancing the defensive capabili-ties of black-box testing and ensuring each metric possesses sufficient orthogonality.In experimental validation,the proposed method demonstrates high efficiency in copyright judgment accuracy and reliability,effectively reducing the number of highly cor-related indicators.

关 键 词：生成式人工智能 深度学习模型 版权保护 黑盒防御 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]