二维码图像安全的理论与关键技术

Due to the rapid development of mobile devices, 2D barcode has been used as a window for the online-to-offline regimes and has been deployed widely in anti-counterfeiting tracing systems as well as mobile payment applications. However, some loopholes of 2D barcode are now threatening the security of the entire mobile network. Within all security issues, the most significant one is how to ensure the image authenticity against replication attack with considerations on the generality of 2D barcode. This project innovates the forensic theory and technology on 2D barcode images against replication attack with consideration on the decoding robustness under low resolution imaging devices. The characteristics of physical devices of the 2D barcode communication channel in the decoding process is utilized in analyzing the resampling artifacts in spatial/frequency feature spaces, and in designing a secure 2D barcoding scheme. A latent feature space will then be derived based on the communication channel model so as to construct a robust identification scheme for 2D barcode under replication attack. The main research contents of the project include: 1) A secure 2D barcoding scheme based on random texture image; 2) A robust feature representation for the 2D barcode resampling process in spatial and frequency domains; 3) A machine learning scheme based on 2D barcode channel model for small training sample size. Based on the above theoretical research, the limitations of existing approach, such as low generality, low resistance towards image recovery attack and the need for additional online information can be mitigated. The security of 2D barcode in anti-counterfeiting tracing system and mobile payment application can then be improved.

随着智能移动终端的迅猛发展，二维码作为线上与线下连接的窗口被广泛应用于防伪溯源及移动支付领域。然而，目前二维码在安全性方面存在着不可忽视的漏洞。现阶段二维码最主要的安全问题是：如何有效抵御针对二维码图像的非法复制攻击并保证其广泛适用性不受到影响。针对此问题，本项目将以抵抗针对二维码图像的非法复制攻击为出发点，设计安全的二维码编码结构，分析合法读取与非法复制的信号重采样操作在时频域的特征，并结合二维码通信信道模型构建不同信道状态所共有的潜特征空间，以在小样本条件下有效识别受到非法复制攻击的二维码。主要研究内容包括：1）抗打印扫描攻击的二维码编码结构设计；2）二维码信号重采样过程的图像时频域特征提取；3）二维码信道模型下小样本机器学习机制的构建。基于对上述理论研究，可缓解现有防伪技术在低通用性、无法抵抗还原攻击与需要额外在线信息方面的局限性，并提高二维码在防伪溯源及移动支付方面的安全性。

本项目以抵抗二维码的非法复制攻击为出发点， 在防复制图案的技术框架下， 研究二维码图像安全的关键理论与技术。为解决现有防复制二维码通用性低、无法抵御非法还原攻击和需要额外在线信息三个方面的局限性，本项目提出了基于二维码打印扫描信道的失真模型与噪声特性，提取非法复制操作中的本质属性特征，设计了具有较高通用性的防复制二维码，构建了在跨域及小样本场景中适用的检测算法。..本项目以抵抗针对二维码图像的非法复制攻击为出发点，设计安全的二维码编码结构，分析合法读取与非法复制的信号重采样操作在时频域的特征，并结合二维码通信信道模型构建不同信道状态所共有的潜特征空间，以在小样本条件下有效识别受到非法复制攻击的二维码。主要研究内容包括：1）抗打印扫描攻击的二维码编码结构设计；2）二维码信号重采样过程的图像时频域特征提取；3）二维码信道模型下小样本机器学习机制的构建。..本项目的重要研究结果为设计了构建抗非法复制及图像还原攻击的安全二维码编码结构设计方法；分析了打印-扫描信道的失真模型，从而对不同环境下获取的特征空间进行域迁移；取得了对信号重采样操作在 LBP 域与二维傅里叶变换域的准确表征，提出二维码的防复制检测方案；构建了实时的信道参数估计理论，从而完成各特征空间的样本聚类；并将小样本机器学习理论从二维码防复制问题有效拓展至人脸翻拍检测问题。..本项目的成果提出并完成了二维码信道建模、防复制特征提取和小样本跨域机器学习算法的研究任务，发表高质量学术论文8篇（含SCI论文7篇，EI论文1篇。IEEE Transactions 系列论文5篇），获得授权中国专利4项，申请PCT国际专利2项，培养3名博士后、1名博士研究生和4名硕士研究生。在本项目研究成果紧贴防伪溯源系统的实际需求，提出了防复制的二维码设计方案及检测机制，有望增强目前防伪溯源系统的安全性，从而有效打击假冒伪劣产品。