Bag of Tricks for Training Data Extraction from Language Models

Weichen Yu, Tianyu Pang, Qian Liu, Chao Du, Bingyi Kang, Yan Huang, Min Lin, Shuicheng Yan
Proceedings of the 40th International Conference on Machine Learning, PMLR 202:40306-40320, 2023.

Abstract

With the advance of language models, privacy protection is receiving more attention. Training data extraction is therefore of great importance, as it can serve as a potential tool to assess privacy leakage. However, due to the difficulty of this task, most of the existing methods are proof-of-concept and still not effective enough. In this paper, we investigate and benchmark tricks for improving training data extraction using a publicly available dataset. Because most existing extraction methods use a pipeline of generating-then-ranking, i.e., generating text candidates as potential training data and then ranking them based on specific criteria, our research focuses on the tricks for both text generation (e.g., sampling strategy) and text ranking (e.g., token-level criteria). The experimental results show that several previously overlooked tricks can be crucial to the success of training data extraction. Based on the GPT-Neo 1.3B evaluation results, our proposed tricks outperform the baseline by a large margin in most cases, providing a much stronger baseline for future research. The code is available at https://github.com/weichen-yu/LM-Extraction.

Cite this Paper

BibTeX
@InProceedings{pmlr-v202-yu23c, title = {Bag of Tricks for Training Data Extraction from Language Models}, author = {Yu, Weichen and Pang, Tianyu and Liu, Qian and Du, Chao and Kang, Bingyi and Huang, Yan and Lin, Min and Yan, Shuicheng}, booktitle = {Proceedings of the 40th International Conference on Machine Learning}, pages = {40306--40320}, year = {2023}, editor = {Krause, Andreas and Brunskill, Emma and Cho, Kyunghyun and Engelhardt, Barbara and Sabato, Sivan and Scarlett, Jonathan}, volume = {202}, series = {Proceedings of Machine Learning Research}, month = {23--29 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v202/yu23c/yu23c.pdf}, url = {https://proceedings.mlr.press/v202/yu23c.html}, abstract = {With the advance of language models, privacy protection is receiving more attention. Training data extraction is therefore of great importance, as it can serve as a potential tool to assess privacy leakage. However, due to the difficulty of this task, most of the existing methods are proof-of-concept and still not effective enough. In this paper, we investigate and benchmark tricks for improving training data extraction using a publicly available dataset. Because most existing extraction methods use a pipeline of generating-then-ranking, i.e., generating text candidates as potential training data and then ranking them based on specific criteria, our research focuses on the tricks for both text generation (e.g., sampling strategy) and text ranking (e.g., token-level criteria). The experimental results show that several previously overlooked tricks can be crucial to the success of training data extraction. Based on the GPT-Neo 1.3B evaluation results, our proposed tricks outperform the baseline by a large margin in most cases, providing a much stronger baseline for future research. The code is available at https://github.com/weichen-yu/LM-Extraction.} }
Endnote
%0 Conference Paper %T Bag of Tricks for Training Data Extraction from Language Models %A Weichen Yu %A Tianyu Pang %A Qian Liu %A Chao Du %A Bingyi Kang %A Yan Huang %A Min Lin %A Shuicheng Yan %B Proceedings of the 40th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2023 %E Andreas Krause %E Emma Brunskill %E Kyunghyun Cho %E Barbara Engelhardt %E Sivan Sabato %E Jonathan Scarlett %F pmlr-v202-yu23c %I PMLR %P 40306--40320 %U https://proceedings.mlr.press/v202/yu23c.html %V 202 %X With the advance of language models, privacy protection is receiving more attention. Training data extraction is therefore of great importance, as it can serve as a potential tool to assess privacy leakage. However, due to the difficulty of this task, most of the existing methods are proof-of-concept and still not effective enough. In this paper, we investigate and benchmark tricks for improving training data extraction using a publicly available dataset. Because most existing extraction methods use a pipeline of generating-then-ranking, i.e., generating text candidates as potential training data and then ranking them based on specific criteria, our research focuses on the tricks for both text generation (e.g., sampling strategy) and text ranking (e.g., token-level criteria). The experimental results show that several previously overlooked tricks can be crucial to the success of training data extraction. Based on the GPT-Neo 1.3B evaluation results, our proposed tricks outperform the baseline by a large margin in most cases, providing a much stronger baseline for future research. The code is available at https://github.com/weichen-yu/LM-Extraction.
APA
Yu, W., Pang, T., Liu, Q., Du, C., Kang, B., Huang, Y., Lin, M. & Yan, S.. (2023). Bag of Tricks for Training Data Extraction from Language Models. Proceedings of the 40th International Conference on Machine Learning, in Proceedings of Machine Learning Research 202:40306-40320 Available from https://proceedings.mlr.press/v202/yu23c.html.

Related Material