Enhancement of Markov chain-based linguistic steganography with binary encoding for securing legal documents

ABSTRACT: This study addresses the limitations of traditional linguistic steganography by enhancing a Markov Chain-based algorithm with binary encoding to improve efficiency, capacity, and reliability. Conventional methods often suffer from slow processing, low embedding capacity, and poor decoding accuracy due to complex encoding mechanisms and lack of optimization, making them unsuitable for high-speed, precise applications. The proposed enhancements remove Huffman tree construction, reducing time complexity to O(1) and achieving up to a 54-fold improvement in encoding and decoding efficiency for large bitstreams. Embedding rates are further optimized through entry point utilization and binary indexing, with Markov models of state sizes 2 and 3 enabling higher data capacities while keeping outputs compact. To ensure robustness, a validation mechanism for incomplete bit groups was introduced, yielding a 100% decoding success rate compared to only 4% in existing methods. Imperceptibility was measured using perplexity, confirming that the algorithm generates human-like text at higher state sizes. These advancements make the algorithm suitable for securely embedding sensitive data in domains such as legal and financial communications. Overall, the enhanced algorithm offers a fast, reliable, and scalable solution for text-based steganography, overcoming the inefficiencies of prior methods and contributing significantly to the theoretical and practical development of secure information hiding techniques.