# GrayShield: Bit-Level Sanitization for Transformer Model Supply-Chain Security

arXiv: https://arxiv.org/abs/2610.04319 · HTML: https://mlsanity.com/paper/2610.04319/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 12 of its 26 references that are in the map.
- First posted on arXiv 3 Oct 2026.

## Builds on (most cited first, 12 of 12)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [RoBERTa: A Robustly Optimized BERT Pretraining Approach](https://mlsanity.com/paper/1907.11692/) (2019)
- [Swin Transformer: Hierarchical Vision Transformer using Shifted Windows](https://mlsanity.com/paper/2103.14030/) (2021)
- [GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding](https://mlsanity.com/paper/1804.07461/) (2018)
- [GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformers](https://mlsanity.com/paper/2210.17323/) (2022)
- [Up or Down? Adaptive Rounding for Post-Training Quantization](https://mlsanity.com/paper/2004.10568/) (2020)
- [Quantizing deep convolutional networks for efficient inference: A whitepaper](https://mlsanity.com/paper/1806.08342/) (2018)
- [DeepHammer: Depleting the Intelligence of Deep Neural Networks through Targeted Chain of Bit Flips](https://mlsanity.com/paper/2003.13746/) (2020)
- [Models Are Codes: Towards Measuring Malicious Code Poisoning Attacks on Pre-trained Model Hubs](https://mlsanity.com/paper/2409.09368/) (2024)
- [EvilModel: Hiding Malware Inside of Neural Network Models](https://mlsanity.com/paper/2107.08590/) (2021)
- [A Large-Scale Exploit Instrumentation Study of AI/ML Supply Chain Attacks in Hugging Face Models](https://mlsanity.com/paper/2410.04490/) (2024)
- [Spectrum-Adaptive Generalization Bounds for Trained Deep Transformers](https://mlsanity.com/paper/2605.07297/) (2026)

## Abstract (arXiv)

Transformer models such as BERT and Vision Transformer~(ViT) achieve strong performance via densely parameterized attention backbones. However, the least significant bits~(LSBs) of their 32-bit floating-point weights can be abused as covert channels to conceal malicious payloads, posing a serious threat to the AI model supply chain. We propose GS (GSabbr), a lightweight, post-training, zero-data sanitization method that completely replaces the declared mantissa-LSB channel with a Gray-code-guided low-transition sequence. Complete payload-independent overwrite, whether keyed or public, makes the sanitized target bits independent of the embedded payload and gives that declared channel zero capacity. Gray coding supplies overwrite structure, while a keyed per-tensor phase supplies pattern diversity. Benchmarked against seven post-training defenses on four Transformer model presets and two real-world malware payloads, GSabbr maintains sub-1% accuracy impact and achieves 49.96±0.66 percentage-point Recovery Reduction (RR) under five implemented attacker variants. Because pre-defense recovery is effectively 100%, RR near 50 percentage points corresponds to post-sanitization bit accuracy at binary chance. Its main empirical advantage is stable near-chance sanitization with substantially smaller weight-distribution shift than the evaluated near-chance baselines PatternMask (PM) and Post-Training Quantization (PTQ).