Model Extraction and Privacy Leakage
Synthesis
Model extraction, privacy leakage, and side channels remain relevant where models expose APIs, retrieval context, memory, or cached computation.
Evidence Base
This page was generated from the batch ingest of SRC-20260702-raw-papers-batch, SRC-20260702-raw-whitepapers-batch, and SRC-20260702-raw-news-batch. Treat it as a navigation and synthesis page; promote individual statements into claim pages when they become decision-relevant.
Representative Sources
| Title | Kind | Date | Tags | Raw |
|---|---|---|---|---|
| Security and Privacy in Retrieval-Augmented Generation: Architectures, Threats, Defenses, and Future | paper | 2026-06-24 | ai-security, defense, privacy, rag, survey, threat-model | raw |
| Anthropic accuses Alibaba of running the largest distillation campaign yet against Claude | news | 2026-06-24 | ai-security, alibaba, claude, model-distillation, model-extraction, qwen | raw |
| Privacy-Preserving RAG via Multi-Agent Semantic Rewriting: Achieving Confidentiality Without Comprom | paper | 2026-06-23 | ai-security, multi-agent-systems, privacy, rag, semantic-rewriting | raw |
| GIF: Locally Sound Geometric Information Flow Control for LLMs | paper | 2026-06-22 | ai-security, data-leakage, formal-methods, information-flow-control, llm | raw |
| Agent-Assisted Side-Channel Attacks on Non-Prefix KV Cache in RAG | paper | 2026-06-20 | agent-assisted-attack, ai-security, kv-cache, rag, side-channel | raw |
| \"What Happens Locally, Leaks Globally\": Detecting Privacy Leakage Risks in MCP Servers | paper | 2026-06-19 | agent-security, ai-security, mcp, privacy-leakage, static-analysis, tool-security | raw |
| From Privacy to Workflow Integrity: Communication-Graph Metadata in Autonomous Agent Interoperabilit | paper | 2026-06-17 | a2a, ai-security, mcp, metadata-leakage, multi-agent-systems, workflow-integrity | raw |
| An Evaluation of Data Leakage Risks in Tool-Using LLM Agents in Realistic Scenarios | paper | 2026-06-15 | agent-security, ai-security, data-leakage, privacy, tool-using-agent | raw |
| Let Them Steal: Trapping Large Language Model Extraction Attacks with Knowledge Honeypot | paper | 2026-06-14 | ai-security, defense, honeypot, llm, model-extraction | raw |
| AI Model Extraction Attacks: Bypassing Single-Client Assumptions in Defenses | paper | 2026-06-02 | Gustavo S찼nchez, Johannes F. Loevenich, Laurin Holz, Maxime Schwarzer, Roberto Rigolin F. Lopes, Thies M철hlenhof | raw |
| An Embarrassingly Simple Detector for Model Extraction Attacks in LLM APIs | paper | 2026-06 | detection, latest-research, llm-api-security, model-extraction, security-for-ai | raw |
| SS-ZKR: Spatial-Semantic Zero-Knowledge Routing for Privacy-Preserving Multi-Agent Collaboration | paper | 2026-05-31 | a2a, ai-security, mcp, multi-agent-systems, privacy, routing | raw |
Open Questions
- Which evidence in this cluster is strongest enough to support a stable claim?
- Which results generalize across models, tools, and deployment settings?
- Which controls have been evaluated under realistic adversarial conditions?