arXiv:2606. 18430v1 Announce Type: new Abstract: Statistical watermarks help organizations attribute large language model (LLM) outputs, yet existing detectors often struggle when watermark signals are weak, texts are repetitive, or watermarks are edited.
The paper demonstrates that N‑gram based code watermarking schemes, widely used to identify machine‑generated code, are ineffective when faced with realistic code obfuscation. By modeling semantics‑preserving transformations as a Markov random walk and introducing the assumption of distribution consistency, the authors prove that obfuscation can drive the failure rate of any detector to nearly 1 minus its false‑positive rate. Extensive experiments across multiple watermarking methods, LLMs, languages, benchmarks, and obfuscators confirm that detectors collapse to near‑random performance (AUROC ≈ 0.5) after obfuscation.
arXiv:2607. 06009v1 Announce Type: cross Abstract: Attributing code to the large language model that produced it is essential for provenance, licensing, and misuse accountability, yet no deployed watermark meets this need.
arXiv:2502. 02068v3 Announce Type: replace-cross Abstract: This paper introduces RoSeMary, the first-of-its-kind ML/Crypto codesign watermarking framework that regulates LLM-generated code to avoid intellectual property rights violations and inappropriate misuse in software development.
The paper introduces an adaptive embedding displacement attack (EDA) that exploits rewording, reordering, and resegmentation to remove semantic watermarks from text, achieving a 32.6%–47.9% success rate across four watermarking schemes. To counter this, the authors propose k‑SwordStamp, a semantic watermarking method that uses order‑robust detection over sub‑sentence units, significantly reducing vulnerability to structure‑based edits. Experiments show that EDA remains effective against k‑SwordStamp, but with a lower success rate (10.8%) compared to its performance on other schemes.
arXiv:2607. 07881v1 Announce Type: cross Abstract: Large language models (LLMs) are increasingly used for code generation, but they struggle to generate functional code free of security vulnerabilities.
The paper investigates whether code large language models (CodeLLMs) inadvertently reproduce proprietary or sensitive code by evaluating seven state‑of‑the‑art training data detection (TDD) methods on eight CodeLLMs. It introduces CodeSnitch, a benchmark of 9,000 function‑level code samples across three languages, each labeled as included or excluded from training data, and applies mutation strategies based on the Type‑1 to Type‑4 code clone taxonomy to test TDD robustness. The study offers a systematic assessment of current TDD techniques for code and suggests directions for developing more effective detection methods.
Behavioral watermarking embeds an owner identifier in an LLM agent's high-level action choices, giving provenance without touching output tokens. Prior agent watermarks break in two ways. First, all t...
arXiv:2606. 00613v1 Announce Type: cross Abstract: Watermarking should identify language-model output without degrading quality or limiting verification to the model provider.
arXiv:2610.08668v1 Announce Type: cross Abstract: Behavioral watermarking embeds an owner identifier in an LLM agent's high-level action choices, giving provenance without touching output tokens. Pri...
arXiv:2609.38722v1 Announce Type: cross Abstract: LLM watermarking has become an effective approach to distinguishing AI-generated text from human-written text by embedding detectable patterns during...
The paper introduces CodeScan, a black-box, vulnerability-oriented scanning framework designed to detect data poisoning and backdoor attacks in code generation large language models (LLMs). CodeScan operates by analyzing structural similarities across multiple code generations, normalizing them with abstract syntax tree (AST) techniques, and then applying LLM-based vulnerability analysis to identify recurring insecure patterns. Evaluations on 117 models across three architectures and multiple sizes show over 97% detection accuracy with fewer false positives compared to prior methods.