New research questions polynomial-time complexity of DNN model extraction

A new research paper challenges the assumption that extracting information from deep neural networks (DNNs) is always a polynomial-time process. While previous work suggested that hard-label extraction attacks, which only reveal the final classification, are efficient, this paper argues that the complexity increases with network depth. The authors propose a novel cross-layer extraction method to overcome these limitations. AI

影响 Challenges the efficiency assumptions for model extraction attacks, potentially impacting intellectual property protection for DNNs.

排序理由 Academic paper published on arXiv discussing theoretical limitations of model extraction attacks on deep neural networks. [lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.LG 阅读 →

• Akira Ito
• Asiacrypt 2024
• Carlini et al.
• Crypto 2020
• Deep Neural Networks
• Eurocrypt 2024
• Eurocrypt 2025
• hard-label extraction
• ReLU

• 论文
• 安全

AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →