机器遗忘综述  

Survey on Machine Unlearning

作　　者：李梓童 孟小峰[1] 王雷霞 郝新丽 LI Zi-Tong;MENG Xiao-Feng;WANG Lei-Xia;HAO Xin-Li(School of Information,Renmin University of China,Beijing 100872,China)

机构地区：[1]中国人民大学信息学院,北京100872

出　　处：《软件学报》2025年第4期1637-1664,共28页Journal of Software

基　　金：国家自然科学基金(61941121,91846204,6217242)。

摘　　要：近年来,机器学习在人们日常生活中应用愈发广泛,这些模型在历史数据上进行训练,预测未来行为,极大地便利了人们生活.然而,机器学习存在隐私泄露隐患:当用户不希望个人数据被使用时,单纯地把其数据从训练集中删去并不够,已训练好的模型仍包含用户信息,可能造成隐私泄露.为了解决这一问题,让机器学习模型“遗忘”该用户个人数据,最简单的方法是在不包含其数据的训练集上重新训练,此时得到的新模型必定不包含个人数据的信息.然而,重新训练往往代价较大,成本较高,由此产生“机器遗忘”的关键问题:能否以更低的代价,获取与重新训练模型尽可能相似的模型.对研究这一问题的文献进行梳理归纳,将已有机器遗忘方法分为基于训练的方法、基于编辑的方法和基于生成的方法这3类,介绍机器遗忘的度量指标,并对已有方法进行测试和评估,最后对机器遗忘作未来展望.Machine learning has become increasingly prevalent in daily life.Various machine learning methods are proposed to utilize historical data for making predictions,making people’s life more convenient.However,there is a significant challenge associated with machine learning-privacy leakage.Mere deletion of a user’s data from the training set is not sufficient for avoiding privacy leakage,as the trained model may still harbor this information.To tackle this challenge,the conventional approach entails retraining the model on a new training set that excludes the data of the user.However,this method can be costly,prompting the exploration for a more efficient way to“unlearn”specific data while yielding a model comparable to a retrained one.This study summarizes the current literature on this topic,categorizing existing unlearning methods into three groups:training-based,editing-based,and generation-based methods.Additionally,various metrics are introduced to assess unlearning methods.The study also evaluates current unlearning methods in deep learning and concludes with future research directions in this field.

关 键 词：机器学习 机器遗忘 深度学习 隐私保护 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]