Lune

CRYPTO2026Top-tier venue

Algebraic Attack on Convolutional Neural Networks with Max Pooling

Zirui Chen, Shi Tang, Zhengchao Gao, Yongjia Su, Lingyue Qin, Xiaoyang Dong

Abstract

Recovering the weights and biases of deep neural networks (DNNs) via black-box input-output queries, known as parameter extraction attacks, has been extensively studied for ReLU-based fully connected neural networks (FCNNs), but remains unexplored for convolutional neural networks (CNNs) with the max pooling function, a core architecture for computer vision and multimedia processing. The key challenge lies in the CNN max pooling layer, which introduces an additional non-linearity and hides ReLU critical points, rendering existing FCNN extraction methods inapplicable. To address this gap, we propose the first cryptanalytic extraction attack tailored for CNNs with the max pooling function. First, we establish an algebraic representation of CNNs, formally proving that CNNs are piecewise linear functions enabling the extension of linearity-based extraction principles. We then identify two novel types of critical points in CNNs: ReLU-Pooling Critical Points (RPCPs) and Pooling Switching Points (PSPs). We design complementary extraction techniques: a pattern matching method for RPCPs to recover partial signatures and signs, and an internal differential extraction attack for PSPs, inspired by cryptographic internal differential analysis, to recover high-accuracy signatures. Given that PSPs are far more abundant than RPCPs and yield a highly efficient extraction method, and that RPCPs are indispensable for bias recovery, we integrate both methods: the PSP method enables efficient signature extraction, while a single RPCP recovers the sign and bias. We evaluate our attack on multiple CNN architectures, including modern adaptations of LeNet-5, trained on random data, MNIST, and CIFAR-10. Experimental results demonstrate that our approach achieves high extraction accuracy with polynomial query complexity and runtime, even for deep CNN layers. This work fills a research gap in CNN security.

Ask Lune

Know this paper in minutes, not an afternoon.

Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.

Questions to start from

Lune can draw on

the full paper · related work

Get the answer

Free to start. No credit card required.

Lineage

  • Beyond Slow Signs in High-fidelity Model Extraction cites Cryptanalytic Extraction of Neural Network Models
  • Beyond Slow Signs in High-fidelity Model Extraction cites Reverse-engineering deep ReLU networks
  • Cryptanalytic Extraction of Neural Network Models cites CSI NN: Reverse Engineering of Neural Network Architectures Through Electromagnetic Side Channel
  • Cryptanalytic Extraction of Neural Network Models cites Reverse-engineering deep ReLU networks
Builds onCited byLeft to right by year; the more cited, the higher.

Builds on10

  1. Oblivious Neural Network Predictions via MiniONN TransformationsJian Liu, Mika Juuti, Yao Lu, N. AsokanCCS 2017 · 800 citations
  2. CSI NN: Reverse Engineering of Neural Network Architectures Through Electromagnetic Side ChannelLejla Batina, Shivam Bhasin, Dirmanto Jap, Stjepan PicekUSENIX Security 2019 · 334 citations
  3. Reverse-engineering deep ReLU networksDavid Rolnick, Konrad P. KordingICML 2020 · 121 citations
  4. Cryptanalytic Extraction of Neural Network ModelsNicholas Carlini, Matthew Jagielski, Ilya MironovCRYPTO 2020 · 109 citations
  5. Beyond Slow Signs in High-fidelity Model ExtractionHanna Foerster, Robert Mullins, Ilia Shumailov, Jamie HayesNeurIPS 2024 · 19 citations

Related papers

  1. Polynomial Time Cryptanalytic Extraction of Deep Neural Networks in the Hard-Label SettingNicholas Carlini, Jorge Chávez-Saab, Anna Hambitzer, Francisco Rodríguez-Henríquez et al.EUROCRYPT 2025 · 10 citations
  2. Navigating the Deep: End-to-End Extraction on Deep Neural NetworksHaolin Liu, Adrien Siproudhis, Samuel Experton, Peter Lorenz et al.EUROCRYPT 2026 · 2 citations
  3. Polynomial Time Cryptanalytic Extraction of Neural Network ModelsIsaac Andrés Canales Martinez, Jorge Chávez-Saab, Anna Hambitzer, Francisco Rodríguez-Henríquez et al.EUROCRYPT 2024 · 13 citations
  4. Cryptanalytic Extraction of Deep Neural Networks with Non-linear ActivationsRoderick Asselineau, Patrick Derbez, Pierre-Alain Fouque, Brice MinaudCRYPTO 2026 · 8 citations
  5. Is the Hard-Label Cryptanalytic Model Extraction Really Polynomial?Akira Ito, Takayuki Miura, Yosuke TodoCRYPTO 2026
Dusk over the sea between two cliffs drawn in fine vertical lines