SparSEEty: Extracting Tokens from Sparsity-Exploiting LLM Serving Systems via Deterministic Side Channels
Yongwan Jo, Jinyoung Park, Euihyun Lee, Dokyung Song
cs.CR, cs.AI
Submitted: 2026-08-04
Code: https://github.com/ggml-org/llama.cpp
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Toy Models of Superposition
- Steering Large Language Model Activations in Sparse Spaces
- PowerInfer-2: Fast Large Language Model Inference on a Smartphone
- LoRA: Low-Rank Adaptation of Large Language Models
- ReLU$^2$ Wins: Discovering Efficient Activation Functions for Sparse LLMs
- Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism
- Llama 2: Open Foundation and Fine-Tuned Chat Models
- Gemma: Open Models Based on Gemini Research and Technology
- MedAlpaca -- An Open-Source Collection of Medical Conversational AI Models and Training Data
- Neural Legal Judgment Prediction in English
- The Early Bird Catches the Leak: Unveiling Timing Side Channels in LLM Serving Systems
- System Prompt Extraction Attacks and Defenses in Large Language Models
- From Persona to Personalization: A Survey on Role-Playing Language Agents
- InputSnatch: Stealing Input in LLM Services via Timing Side-Channel Attacks
- Buffer Overflow in Mixture of Experts
- Stealing User Prompts from Mixture of Experts
- When Speculation Spills Secrets: Side Channels via Speculative Decoding In LLMs
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
- Estimating the Social Cost of Corporate Data Breaches
- Efficient Fuzzy PSI under One-Sided Assumptions
- Sealing the Audit-Runtime Gap for LLM Skills