HookSteg: A Truly Lossless Steganography Based on Process Hooking for Lossy Network Channels

Covert communication often relies on image steganography to transmit sensitive data such as commercial secrets. Any minor bit error can lead to catastrophic consequences, which imposes extremely high demands on the zero-error transmission of secret information. However, existing robust image steganographic schemes still fail to guarantee zero-error recovery of secret bits when resisting hybrid attacks such as compression and scaling on social platforms. To address this issue, we propose a lossless steganographic method called HookSteg, which bypasses lossy attacks on social platforms through process hooking, thereby achieving lossless steganography. This method first identifies nodes affected by lossy attacks through tracking whether image data has been compromised. Subsequently, a method for capturing lossy attack instructions is proposed based on syntax-semantic joint features and dynamic taint tracking techniques. Finally, process hooking is utilized to bypass lossy attack instructions, ensuring the lossless transmission of secret information. Experimental results on transmission scenarios with tens of millions of daily active users, such as Youku and Kutumb, demonstrate that HookSteg is general and requires no modifications to existing steganographic algorithms. It achieves, for the first time, lossless transmission of secret information in real-world scenarios, with an average improvement in extraction accuracy of 36.63%. Additionally, its efficiency in locating attack instructions is improved by 94.34% compared to traditional manual reverse engineering methods. Furthermore, this study resolves the trade-off among embedding capacity, security, and robustness, enabling steganography to focus on enhancing security and capacity, thereby providing a general and efficient solution for lossless steganography.