一篇新论文探讨了在模型所有者和审计员之间相互不信任的情况下，审计机器学习遗忘（MU）所面临的挑战。该研究提供了信息论证明，表明通用的行为审计无法在不泄露有关保留数据敏感信息的情况下识别遗忘不充分的模型。这种固有的隐私-审计权衡即使在非凸模型中也存在，表明需要更强大的隐私保护审计方案。 AI
影响 强调了审计AI模型的基本张力，可能影响隐私保护AI系统的开发。
排序理由 该集群包含一篇在arXiv上发表的研究论文，详细介绍了机器学习遗忘的理论和实证研究结果。
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
arXiv:2606.16110v1 Announce Type: new Abstract: Machine unlearning has been extensively studied in response to growing privacy concerns and regulatory requirements. However, auditing whether unlearning algorithms have truly erased the influence of specific data remains an open ch…
arXiv:2606.14518v1 Announce Type: new Abstract: The removal of learned data from Machine Learning models through Machine Unlearning (MU) has been widely studied; however, there has yet to be an agreed-upon scheme for auditing MU. Existing work has shown that a dishonest model own…
The removal of learned data from Machine Learning models through Machine Unlearning (MU) has been widely studied; however, there has yet to be an agreed-upon scheme for auditing MU. Existing work has shown that a dishonest model owner can falsify evidence to avoid executing MU, w…