arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 1811.00513cs.CRcs.CLcs.LGstat.ML
收藏

文本生成模型中的审计数据来源

Auditing Data Provenance in Text-Generation Models

Cornell University(康奈尔大学);Cornell Tech(康奈尔科技校区)

Congzheng Song, Vitaly Shmatikov

摘要:

为了帮助执行GDPR等数据保护法规并检测对个人数据的未经授权使用,我们开发了一种新的\emph{model auditing}技术,可以帮助用户检查他们的数据是否被用于训练机器学习模型。我们专注于审计生成自然语言文本的深度学习模型,包括单词预测和对话生成。这些模型是流行的在线服务的核心,通常是根据用户的消息、搜索、聊天和评论等个人数据进行训练的。我们设计和评估一个黑盒审计方法,可以检测到,很少的查询模型,如果一个特定的用户的文本被用来训练它(在成千上万的其他用户)。我们的经验表明,我们的方法可以成功地审计良好的广义模型,没有过拟合的训练数据。我们还分析了文本生成模型如何记忆单词序列,并解释了为什么这种记忆使他们适合审计。

英文摘要:

To help enforce data-protection regulations such as GDPR and detect unauthorized uses of personal data, we develop a new \emph{model auditing} technique that helps users check if their data was used to train a machine learning model. We focus on auditing deep-learning models that generate natural-language text, including word prediction and dialog generation. These models are at the core of popular online services and are often trained on personal data such as users' messages, searches, chats, and comments. We design and evaluate a black-box auditing method that can detect, with very few queries to a model, if a particular user's texts were used to train it (among thousands of other users). We empirically show that our method can successfully audit well-generalized models that are not overfitted to the training data. We also analyze how text-generation models memorize word sequences and explain why this memorization makes them amenable to auditing.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑