Chinese Text Information Hiding Based On Character Perturbation Deformation And Controllable Generationon

Text information hiding refers to hiding the information in the text carrier in a way that is not easy to detect,so as to ensure the safe transmission of information on the open channel.As text is the most widely used medium for information transmission,information hiding technology using text as a carrier also holds a wide range of application prospects.Text information hiding technology mainly contains text digital watermarking technology and text steganography.In recent years there have been new advances in text digital watermarking technology,a text digital watermarking method based on glyph design and character replacement,which has good performance in watermark capacity and visual quality.However,the current method still suffers from complex and difficult Chinese character design and generation,and low watermarking robustness.On the other hand,a class of generative linguistic steganography has emerged in the study of steganography for text.This type of method uses natural language generation techniques to automatically generate steganographic text based on secret messages,breaking the limit on the length of the text carrier and therefore achieving a higher hiding capacity.At the same time,there are other challenges,such as uncontrollable semantics,imperceptibility and security of the steganographic text,which require further research to achieve usability.To address these issues,this paper focuses on digital watermarking techniques and generative steganography for Chinese text,with the main contents as follows:1)A robust Chinese text watermarking method based on Chinese character glyph perturbation and font replacing is proposed.The method generates a perturbed and deformated Chinese characters font by modifying the stroke structure features of Chinese characters,and numbers different perturbed glygh of the same Chinese character to achieve watermark information embedding.Based on the frequency of use of Chinese characters and the correlation between characters,a grouping algorithm of perturbed characters glyph is designed to improve the efficiency and performance of watermark embedding and extraction.Simulation and ablation experiments are conducted to verify the feasibility of the method,and comparative analysis and robustness tests are used to demonstrate the performance of the method.2)A Chinese text generation steganography method with controllable content and security is proposed.The method uses a controlled generative language model to generate steganographic text and designs a dynamic threshold sampling strategy to ensure the security of the steganographic text.To make the model more suitable for steganography generation tasks,a reinforcement learning-based policy optimisation is used to fine-tune the generative model,and encourage the model to generate more secure and high-quality steganographic text.The method is fine-tuned and experimentally analysed on an open-source question-and-answer dataset.In terms of objective metrics,the experimental results show that the method has a high hiding capacity and good performance.In terms of subjective evaluation,the generated high quality steganographic text also receives high recognition.In terms of steganalysis resistance,the steganographic text generated by this method is more difficult to be detected by popular classification networks than other existing methods,and it is more secure.