Hollow-LLM Attack: Computationally Trivial Weights in Zero-Knowledge Verification of LLM Inference
Chen Gong, Beijie Liu, Mengyuan Li
cs.CR
Submitted: 2026-07-30
Comments: Accepted to the 2026 IEEE Symposium on Security and Privacy (S&P 2026)
DOI: 10.1109/SP63933.2026.00258
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- BackdoorLLM: A Comprehensive Benchmark for Backdoor Attacks and Defenses on Large Language Models
- CoIn: Counting the Invisible Reasoning Tokens in Commercial Opaque LLM APIs
- Are You Getting What You Pay For? Auditing Model Substitution in LLM APIs
- Auditing Black-Box LLM APIs with a Rank-Based Uniformity Test
- Trust, but verify
- Slalom: Fast, Verifiable and Private Execution of Neural Networks in Trusted Hardware
- Chiron: Privacy-preserving Machine Learning as a Service
- FedML-HE: An Efficient Homomorphic-Encryption-Based Privacy-Preserving Federated Learning System
- ZKTorch: Compiling ML Inference to Zero-Knowledge Proofs via Parallel Proof Accumulation
- ezDPS: An Efficient and Zero-Knowledge Machine Learning Inference Pipeline
- SVIP: Towards Verifiable Inference of Open-source Large Language Models
- Distilling the Knowledge in a Neural Network
- DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter
- VeriLLM: A Lightweight Framework for Publicly Verifiable Decentralized Inference
Related papers
- SoK: AI-Augmented Binary Reversing
- Relaxed Sender Anonymity for CBDC Interbank Settlement: A Zero-Knowledge Approach on Permissioned EVM
- Calibration-Family Overfit: Why Trusted Sabotage Monitors Don't Transfer Across Lineages
- Efficient Fuzzy PSI under One-Sided Assumptions
- Sealing the Audit-Runtime Gap for LLM Skills
- Token Composition: A Graph Based on EVM Logs