Abstract:Linguistic steganography provides convenient implementation to hide messages, particularly with the emergence of AI generation technology. The potential abuse of this technology raises security concerns within societies, calling for powerful linguistic steganalysis to detect carrier containing steganographic messages. Existing methods are limited to finding distribution differences between steganographic texts and normal texts from the aspect of symbolic statistics. However, the distribution differences of both kinds of texts are hard to build precisely, which heavily hurts the detection ability of the existing methods in realistic scenarios. To seek a feasible way to construct practical steganalysis in real world, this paper propose to employ human-like text processing abilities of large language models (LLMs) to realize the difference from the aspect of human perception, addition to traditional statistic aspect. Specifically, we systematically investigate the performance of LLMs in this task by modeling it as a generative paradigm, instead of traditional classification paradigm. Extensive experiment results reveal that generative LLMs exhibit significant advantages in linguistic steganalysis and demonstrate performance trends distinct from traditional approaches. Results also reveal that LLMs outperform existing baselines by a wide margin, and the domain-agnostic ability of LLMs makes it possible to train a generic steganalysis model (Both codes and trained models are openly available in

Least significant bit steganography detection with machine learning techniques

JPEG Steganalysis Based on Co-occurrence Features and Ensemble Multiple Hyperspheres OC-SVM

Detection of audio-to-image audio steganography based on peak frequency feature

A New Steganalysis Approach Based on Both Complexity Estimate and Statistical Filter

Steganalysis of LSB matching based on the sum features of average co-occurrence matrix using image estimation

Feature-Based Steganalysis for JPEG Images

A co-occurrence matrix based approach to detect jpeg steganography

Steganalysis of model based steganography and steghide in grayscale JPEG images

A Blind Steganalytic Scheme Based on DCT and Spatial Domain for JPEG Images.

Blind detection for JPEG steganography

Variable Characteristics Based Blind Detection of Hidden Information

Applying Machine Learning to Chi-square Steganalysis: A Case Study

Detecting LSB Matching Steganography in Grayscale Images by Changing Rates of Embedding Impact

Image classification method of distinguishing LSB replacement from matching steganography

Steganalysis Based on Contamination Data Analysis

Steganalysis of LSB Replacement Based on Wavelet Transform

A Novel LSB Steganography Scheme Against Statistical Analysis

A LSB Substitution Compatible Steganography

Low Dimensional Multi Class Steganalysis of Spatial LSB based Stego Images Using Textural Features

A Detection Method of Subliminal Channel Based on VoIP Communication.

Towards Next-Generation Steganalysis: LLMs Unleash the Power of Detecting Steganography