# distillation
標記為「distillation」的 27 篇文章
Knowledge Distillation Attacks
Attacking knowledge distillation pipelines to transfer backdoors from teacher to student models or extract proprietary capabilities.
API-Based Model Extraction
Deep dive into extracting proprietary model capabilities through systematic API querying, active learning strategies, logprob exploitation, soft-label distillation, and evasion of query anomaly detection systems.
Model Extraction & IP Theft
Methodology for black-box model extraction, API-based distillation, side-channel extraction, watermark removal, and model fingerprinting bypass targeting deployed AI systems.
Model Distillation Security Implications
Security risks of knowledge distillation including capability transfer, safety property loss, and unauthorized model cloning.
Safety Loss During Model Distillation
Research on how safety alignment degrades during knowledge distillation from larger to smaller models.
Model Compression Security
Security implications of model pruning, quantization, and knowledge distillation on AI system robustness.
Model Distillation Security Lab
Extract model capabilities through distillation techniques using only black-box API access.
Inference Optimization Risks
Security implications of model optimization techniques — covering quantization safety degradation, pruning vulnerability introduction, distillation attacks, and speculative decoding risks.
Distillation Security Analysis
Security implications of knowledge distillation including backdoor transfer, capability extraction, and safety property degradation in student models.
Model Distillation Attacks
Stealing model capabilities via knowledge distillation: API-based distillation, bypassing access restrictions, task-specific capability theft, and defense against distillation-based model stealing.
Distillation-Based Model Extraction
Using knowledge distillation for model theft: student-teacher extraction attacks, API-based distillation, task-specific extraction, and defending against distillation-based model stealing.
Knowledge Distillation Safety Gap
Analysis of safety property loss during knowledge distillation from teacher to student models.
Knowledge Distillation Security
Security implications of knowledge distillation including capability extraction and safety alignment transfer.
Knowledge Distillation 攻擊s
攻擊ing knowledge distillation pipelines to transfer backdoors from teacher to student models or extract proprietary capabilities.
API-Based 模型 Extraction
Deep dive into extracting proprietary model capabilities through systematic API querying, active learning strategies, logprob exploitation, soft-label distillation, and evasion of query anomaly detection systems.
模型 Extraction & IP Theft
Methodology for black-box model extraction, API-based distillation, side-channel extraction, watermark removal, and model fingerprinting bypass targeting deployed AI systems.
模型 Distillation 安全 Implications
安全 risks of knowledge distillation including capability transfer, safety property loss, and unauthorized model cloning.
Safety Loss During 模型 Distillation
Research on how safety alignment degrades during knowledge distillation from larger to smaller models.
模型 Compression 安全
安全 implications of model pruning, quantization, and knowledge distillation on AI system robustness.
模型 Distillation 安全 實驗室
Extract model capabilities through distillation techniques using only black-box API access.
推論最佳化風險
模型最佳化技術的安全意涵——涵蓋量化安全降級、剪枝漏洞引入、蒸餾攻擊與推測解碼風險。
Distillation 安全 Analysis
安全 implications of knowledge distillation including backdoor transfer, capability extraction, and safety property degradation in student models.
模型 Distillation 攻擊s
Stealing model capabilities via knowledge distillation: API-based distillation, bypassing access restrictions, task-specific capability theft, and defense against distillation-based model stealing.
基於蒸餾的模型擷取
以知識蒸餾進行模型竊取：師生擷取攻擊、以 API 為基礎的蒸餾、任務特定擷取，以及對抗蒸餾式模型竊取的防禦。
架構層級攻擊
鎖定模型架構最佳化的攻擊——涵蓋量化利用、蒸餾攻擊、KV 快取攻擊、MoE 路由操控與上下文視窗利用。
Knowledge Distillation Safety Gap
Analysis of safety property loss during knowledge distillation from teacher to student models.
Knowledge Distillation 安全
安全 implications of knowledge distillation including capability extraction and safety alignment transfer.