Addressing Segmentation Ambiguity in Neural Linguistic Steganography

TL;DR

This paper investigates how segmentation ambiguity affects neural linguistic steganography, causing decoding failures, and proposes simple solutions applicable across languages, including those without explicit word boundaries.

Contribution

It highlights the impact of segmentation ambiguity on decoding success and introduces effective tricks to mitigate this issue in neural linguistic steganography.

Findings

Segmentation ambiguity causes decoding failures in neural linguistic steganography.

Proposed tricks effectively reduce decoding errors across languages.

Solutions are applicable even to languages without explicit word boundaries.

Abstract

Previous studies on neural linguistic steganography, except Ueoka et al. (2021), overlook the fact that the sender must detokenize cover texts to avoid arousing the eavesdropper's suspicion. In this paper, we demonstrate that segmentation ambiguity indeed causes occasional decoding failures at the receiver's side. With the near-ubiquity of subwords, this problem now affects any language. We propose simple tricks to overcome this problem, which are even applicable to languages without explicit word boundaries.