paper-with-me

Papers

Distilling the Undistillable: Learning from a Nasty Teacher

2022-10-21 · Surgan Jandial, Yash Khasbage, Arghya Pal, Vineeth N Balasubramanian, Balaji Krishnamurthy

The inadvertent stealing of private/sensitive information using Knowledge Distillation (KD) has been getting significant attention recently and has guided subsequent defense efforts considering its critical nature. Recent work Nasty Teacher proposed to develop teachers which can not be distilled or imitated by models attacking it. However, the promise of confidentiality offered by a nasty teacher is not well studied, and as a further step to strengthen against such loopholes, we attempt to bypass its defense and steal (or extract) information in its presence successfully. Specifically, we analyze Nasty Teacher from two different directions and subsequently leverage them carefully to develop simple yet efficient methodologies, named as HTC and SCM, which increase the learning from Nasty Teacher by upto 68.63% on standard datasets. Additionally, we also explore an improvised defense method based on our insights of stealing. Our detailed set of experiments and ablations on diverse models/settings demonstrate the efficacy of our approach.

📄 PDF Abstract BibTeX arXiv:2210.11728

Code (1)

surgan12/nastyattacks 공식 구현

Tasks

Knowledge Distillation

Methods 이 논문이 사용한 방법론

RPN A Region Proposal Network, or RPN, is a fully convolutional network that simultaneously predicts object bounds and objectness scores at each position. The RPN is trained…
RoIAlign Region of Interest Align, or RoIAlign, is an operation for extracting a small feature map from each RoI in detection and segmentation based tasks. It removes the harsh…
1x1 Convolution A 1 x 1 Convolution is a convolution with some special properties in that it can be used for dimensionality reduction,…
FPN 설명 없음
Convolution A convolution is a type of matrix operation, consisting of a kernel, a small matrix of weights, that slides over input data performing element-wise multiplication with the…
HTC 설명 없음
Knowledge Distillation A very simple way to improve the performance of almost any machine learning algorithm is to train many different models on the same data and then to average their predictions.…

Similar Papers 제목 키워드 기반

Analyzing the Confidentiality of Undistillable Teachers in Knowledge Distillation

2021-12-01 · NeurIPS 2021 12 · Souvik Kundu, Qirui Sun, Yao Fu, Massoud Pedram 외

Knowledge distillation (KD) has recently been identified as a method that can unintentionally leak private information regarding the details of a teacher model to an unauthorized student. Recent research in developing un…

Knowledge Distillation

Undistillable: Making A Nasty Teacher That CANNOT teach students

2021-05-16 · ICLR 2021 1 · Haoyu Ma, Tianlong Chen, Ting-Kuei Hu, Chenyu You 외

Knowledge Distillation (KD) is a widely used technique to transfer knowledge from pre-trained teacher models to (usually more lightweight) student models. However, in certain situations, this technique is more of a curse…

Knowledge Distillation

Stingy Teacher: Sparse Logits Suffice to Fail Knowledge Distillation

2021-09-29 · Haoyu Ma, Yifan Huang, Tianlong Chen, Hao Tang 외

Knowledge distillation (KD) aims to transfer the discrimination power of pre-trained teacher models to (more lightweight) student models. However, KD also poses the risk of intellectual properties (IPs) leakage of teache…

Knowledge Distillation

Distillation Traps and Guards: A Calibration Knob for LLM Distillability

2026-04-21 · Weixiao Zhan, Yongcheng Jing, Leszek Rutkowski, Dacheng Tao arxiv

Knowledge distillation (KD) transfers capabilities from large language models (LLMs) to smaller students, yet it can fail unpredictably and also underpins model leakage risks. Our analysis revealed several distillation t…

Knowledge Distillation

Pocket Foundation Models: Distilling TFMs into CPU-Ready Gradient-Boosted Trees

2026-05-18 · Aditya Tanna, Nassim Bouarour, Mohamed Bouadi, Vinay kumar Sankarapu 외 arxiv

A fraud scorer needs to answer in under 2 ms. The best tabular foundation models (TFMs) take 151-1,275 ms on GPU. We close this gap by distilling the TFM offline into an XGBoost or CatBoost student that runs natively on …