REED: Post-Training Representation Editing for Cross-Domain Linguistic Steganalysis

2026-05-27

AI Summary

This work addresses the significant performance degradation in cross-domain linguistic steganalysis when test texts originate from unseen domains. To tackle this challenge, the authors propose a post-training representation editing strategy that requires neither architectural modifications nor parameter updates. Building upon a frozen pre-trained steganalyzer, the method achieves cross-domain adaptation by constructing domain-shift vectors or enhances cross-domain generalization through sample-specific editing guided by direction vectors derived from plaintext-to-stegotext mappings in the source domain. Evaluated across multiple cross-domain scenarios, the approach substantially improves F1-scores compared to state-of-the-art methods, offering both deployment flexibility and high detection accuracy without retraining the underlying model.

Abstract

In real-world scenarios of linguistic steganalysis, tested texts usually come from unseen domains with different vocabularies, topics, writing styles, and steganographic generation patterns, which can significantly degrade the detection performance. Although existing cross-domain steganalysis methods can effectively alleviate this problem through distribution alignment, domain-invariant feature learning, etc., the detection performance is not satisfactory. In this paper, we propose a post-training representation editing method for cross-domain linguistic steganalysis. Specifically, the detector is first trained on source-domain data, and then the feature extractor and classifier are kept frozen, and the intermediate representations are deterministically edited before classification. For domain adaptation, we construct a domain-offset vector from marginal source and target representations. For domain generalization, we derive a source-domain cover-to-stego direction to guide sample-specific editing. Experimental results show that compared with the advanced methods, the proposed method can achieve high cross-domain detection performance, especially in terms of F1-score, while requiring no architecture modification or parameter updates after source-domain training.

Problem

• cross-domain
• linguistic steganalysis
• domain shift
• representation editing
• steganographic detection

Innovation

• post-training representation editing
• cross-domain steganalysis
• domain adaptation
• domain generalization
• linguistic steganalysis