mlsanityData to 7 Oct 2026

Seed Hijacking of LLM Sampling and Quantum Random Number Defense

arXiv 2605.08313 ↗First posted 8 May 2026cs.CR cs.AI cs.LGcited by 1

Not on the map. Why?

Abstract (arXiv)

Large language models (LLMs) rely on deterministic pseudorandom number generators (PRNGs) for autoregressive sampling, creating a critical supply-chain attack surface overlooked by existing defenses. We present SeedHijack, a backdoor attack that manipulates PRNG outputs to force attacker-specified token selection without altering model logits. In a 540-trial benchmark on GPT-2 (124M), the attack achieves 99.6% exact token injection across 9 sampling configurations; it reaches 100% success on four aligned models (1.5B-7B, RLHF/SFT/reasoning distillation) and bypasses all alignment methods tested in this work. We further propose a defense based on a hardware quantum random number generator (QRNG), which neutralizes the attack in our evaluated threat model with negligible median overhead (+0.6% latency, +7.7 MB memory). Our work identifies a critical sampling-layer vulnerability and provides a practical, deployable QRNG-based defense.

Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 16 of its 28 references that are in the map. First posted on arXiv 8 May 2026.

Builds on

16 of 28 references.

References · 16

  1. 2022Training language models to follow instructions with human feedbackcited by 15,712
  2. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 7,916
  3. 2014Explaining and Harnessing Adversarial Examplescited by 7,287
  4. 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,326
  5. 2019The Curious Case of Neural Text Degenerationcited by 2,881
  6. 2019Certified Adversarial Robustness via Randomized Smoothingcited by 1,324
  7. 2023BeaverTails: Towards Improved Safety Alignment of LLM via a Human-Preference Datasetcited by 828
  8. 2017Provable defenses against adversarial examples via the convex outer adversarial polytopecited by 564
  9. 2024JailbreakBench: An Open Robustness Benchmark for Jailbreaking Large Language Modelscited by 494
  10. 2018Certified Defenses against Adversarial Examplescited by 305
  11. 2017Certified Defenses for Data Poisoning Attackscited by 288
  12. 2024Bag of Tricks: Benchmarking of Jailbreak Attacks on LLMscited by 47
4 more
  1. 2023Conformal Nucleus Samplingcited by 32
  2. 2025FC-Attack: Jailbreaking Multimodal Large Language Models via Auto-Generated Flowchartscited by 15
  3. 2024Stealthy Jailbreak Attacks on Large Language Models via Benign Data Mirroringcited by 10
  4. 2023Machine Learning needs Better Randomness Standards: Randomised Smoothing and PRNG-based attackscited by 6

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026Blind PRNG Hijacking: An Undetectable Integrity-Preserving Attack Against LLM Watermarkingcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026DiffusionHijack: Supply-Chain PRNG Backdoor Attack on Diffusion Models and Quantum Random Number Defense
  2. 2024Root Defence Strategies: Ensuring Safety of LLM at the Decoding Level
  3. 2025Beyond the Benchmark: Innovative Defenses Against Prompt Injection Attacks
  4. 2025On Jailbreaking Quantized Language Models Through Fault Injection Attacks
  5. 2026The Joint Effect of Quantization and Sampling Temperature on LLM Safety Alignment: A Factorial Analysis
  6. 2025Token-Efficient Prompt Injection Attack: Provoking Cessation in LLM Reasoning via Adaptive Token Compression
  7. 2026AGENTQ: Quantization-Conditioned Backdoor Attacks on LLM Agents
  8. 2026Controlled Decoding Attacks on Black-Box LLMs