Steganography of Steganographic Networks

• DOI: 10.1609/aaai.v37i4.25647
• Publication Date: 2023-06-27T16:43:57Z
• Authors: Guobiao Li; Sheng Li; Meiling Li; Xinpeng Zhang; Zhenxing Qian
• Keywords: Covert; Steganalysis; Steganography tools

Steganography is a technique for covert communication between two parties. With the rapid development of deep neural networks (DNN), more and steganographic are proposed recently, which shown to be promising achieve good performance. Unlike traditional handcrafted tools, network relatively large in size. It raises concerns on how covertly transmit public channels, crucial stage pipeline steganography real world applications. To address such an issue, we propose novel scheme this paper. existing schemes focus subtle modification cover data accommodate secrets. We disguise (termed as secret DNN model) into stego model performs ordinary machine learning task task). During disguising, select tune subset filters preserve its function task, where remaining reactivated according partial optimization strategy whole model. The can recovered from when needed. Various experiments have been conducted demonstrate advantage our method well general models.