Paper says power analysis struggles to extract full model weights
Even a noiseless leakage adversary needs many traces, and deep-layer recovery suffers as models reduce entropy.
Stories per day · 30 days
3 RSS
Relatedbishield-tee1cifar-101compaction1explanation-guidance1fine-tuning1ghostcompact1gpt-5.6-sol1hidden-reasoning1intellectual-property1llm1lora1model-distillation1neural-networks1on-device1openai1parameter-recovery1power-analysis1relu1
22 stories, newest first · one card per story
Stories tagged #model-extraction
VeilTune fine-tunes LoRA adapters entirely inside secure MPC
From one 32-sample mini-batch of hidden states, authors copied a provider adapter on 99.2% of test inputs.
Cutpurse improves model extraction on tabular data
Cutpurse achieves 16.3 percentage point fidelity gain on continuous features without explanation APIs.
OpenAI compaction flaw let researchers extract hidden model reasoning
Xint says OpenAI fixed the server-side issue on October 23 and paid a $6,000 reward; API users need no changes.
1
Paper recovers neural-network parameters by extraction and distillation
On the largest synthetic net, Leaky ReLU recovered 51-57 deepest-layer neurons while ReLU recovered none.
BiShield-TEE: On the (In-)Security of Unilateral Weight Obfuscation in On-Device TEE-Shielded LLM Partition
Unilateral TEE weight obfuscation for on-device LLMs is breakable; BiShield-TEE hides the key among ambiguous solutions.
The Limits of Query Monitoring against Distributed Model Extraction
Study finds query-only monitors detect some distributed model-extraction campaigns but miss slow, low-share, or relayed ones.
MARCO: The Radioactive Watermark for Protein Generative Models
MARCO watermarks protein generative models so marks survive extraction and can trace misused designs.
OpenAI reveals ‘novel’ encryption bypass used in distillation attack
OpenAI disrupted a Moonshot AI-linked distillation campaign that tricked its models into decrypting protected reasoning data using up to 15,000 accounts.
3
Disrupting a coordinated model-distillation campaign
OpenAI disrupted a July campaign extracting protected model reasoning, linking a core cluster to Moonshot AI.
6
Distillation Defenses Easily Break After Reinforcement Learning
Distillation defenses for closed-source LLMs can fail after later reinforcement learning.
Beyond Scalar Probes: Exploiting Vector-Valued Outputs in ReLU Networks For Signature Extraction
Vector-valued Jacobian analysis improves cryptanalytic signature extraction from ReLU networks at low precision.
Distillation Defenses Easily Break After Reinforcement Learning
Distillation defenses for closed language models often fail after later reinforcement learning.
Extracting CNNs in the Unknown-Architecture and Feedback-Agnostic Setting
Researchers show black-box attacks can recover a CNN's unknown architecture and its parameters.
End-to-End Hard-Label Cryptanalytic Model Extraction Using Efficient Sign Recovery
Researchers show end-to-end hard-label extraction of small ReLU MLPs, exceeding 98% label agreement on MNIST.
Normal Alignment: Improved Cryptanalytic Sign Recovery on Hard-Label Networks
Researchers propose Normal Alignment, improving cryptanalytic sign recovery for hard-label neural networks and enabling polynomial-time full model extraction.
RAPID: A Real-Time Defense Against Unauthorized Model Distillation for Text-to-Image Services
RAPID embeds defensive perturbations in a T2I model's shared VAE decoder to block unauthorized black-box distillation in real time.
Shuffling is Not Enough: Breaking Permutation-Based Model Confidentiality in Hybrid FHE Inference
Attack breaks permutation-based model confidentiality in hybrid FHE inference, recovering all ResNet-20 linear layers exactly with d+1 queries per layer.
US Government Accuses Chinese AI Firms of Distilling Frontier Models
US agencies allege Chinese AI firms covertly extracted billions of tokens from OpenAI, Anthropic, Google Gemini, and Grok models to cut development costs.
US Agencies Warn Chinese AI Firms Are Extracting Advanced AI Models
NSA, CISA, and FBI accuse six Chinese AI firms including DeepSeek and Alibaba of industrial-scale distillation of US frontier models.
Understanding the Security Boundary of Obfuscation-based On-Device LLM Protection
Researchers formalize obfuscation primitives for TEE-protected on-device LLMs and show a Collapse attack breaks ArrowCloak, TSQP, and LoRO, then extend the boundary.
CISA Warns Chinese AI Firms Extract Billions of Tokens From Claude, GPT, Gemini and Grok
CISA, NSA and FBI advisory says six Chinese AI firms extracted billions of tokens from Claude, GPT, Gemini and Grok via API proxies since late 2024.