《電子技術(shù)應(yīng)用》
您所在的位置:首頁(yè) > 人工智能 > 设计应用 > 基于黑盒测试框架的深度学习模型版权保护方法
基于黑盒测试框架的深度学习模型版权保护方法
网络安全与数据治理
屈详颜1,2,于静1,2,熊刚1,2,盖珂珂3
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
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。
中圖分類號(hào):TP181
文獻(xiàn)標(biāo)識(shí)碼:ADOI:10.19358/j.issn.2097-1788.2023.12.001
引用格式:屈詳顏,于靜,熊剛,等.基于黑盒測(cè)試框架的深度學(xué)習(xí)模型版權(quán)保護(hù)方法[J].網(wǎng)絡(luò)安全與數(shù)據(jù)治理,2023,42(12):1-6,13.
Copyright protection for deep learning models utilizing a black box testing framework
Qu Xiangyan 1,2, Yu Jing1,2, Xiong Gang1,2, Gai Keke3
1 Institute of Information Engineering, Chinese Academy of Sciences, Beijing 100085, China; 2 School of Cyber Security, University of Chinese Academy of Sciences, Beijing 100049, China; 3 School of Cyberspace Science and Technology, Beijing Institute of Technology, Beijing 100081, China
Abstract: 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 generation 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 adversarial attacks. Additionally, new test metrics and algorithms are introduced for blackbox scenarios, enhancing the defensive capabilities 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 correlated indicators.
Key words : generative artificial intelligence; deep learning models; copyright protection; black box defense

引言

在當(dāng)前生成式人工智能技術(shù)的迅猛發(fā)展推動(dòng)下,深度學(xué)習(xí)模型版權(quán)保護(hù)問題日益受到關(guān)注。深度學(xué)習(xí)模型,尤其是大規(guī)模和高性能的模型,因其昂貴的訓(xùn)練成本,容易遭受未授權(quán)的復(fù)制或再現(xiàn),導(dǎo)致版權(quán)侵犯和模型所有者的經(jīng)濟(jì)損失[1-2]。傳統(tǒng)的版權(quán)保護(hù)方法大多依賴于水印技術(shù)[3-4],通過在模型中嵌入特定的水印來確認(rèn)所有權(quán)。盡管這類方法可以提供確切的所有權(quán)驗(yàn)證,但它們對(duì)原有模型具有侵入性,可能會(huì)影響模型性能或引入新的安全風(fēng)險(xiǎn);并且這些方法對(duì)適應(yīng)性攻擊和新興的模型提取攻擊的魯棒性不足[5-6]。


作者信息

屈詳顏1,2,于靜1,2,熊剛1,2,蓋珂珂3

(1 中國(guó)科學(xué)院信息工程研究所,北京100085;2 中國(guó)科學(xué)院大學(xué)網(wǎng)絡(luò)空間安全學(xué)院,北京100049;

3 北京理工大學(xué)網(wǎng)絡(luò)空間安全學(xué)院,北京100081)


文章下載地址:http://ihrv.cn/resource/share/2000005869


weidian.jpg

此內(nèi)容為AET網(wǎng)站原創(chuàng),未經(jīng)授權(quán)禁止轉(zhuǎn)載。