mlsanityData to 14 Oct 2026

I Know What You See: Power Side-Channel Attack on Convolutional Neural Network Accelerators

arXiv 1803.05847 ↗First posted 5 Mar 2018cs.CV cs.LGcited by 27

Not on the map. Why?

Abstract (arXiv)

Deep learning has become the de-facto computational paradigm for various kinds of perception problems, including many privacy-sensitive applications such as online medical image analysis. No doubt to say, the data privacy of these deep learning systems is a serious concern. Different from previous research focusing on exploiting privacy leakage from deep learning models, in this paper, we present the first attack on the implementation of deep learning models. To be specific, we perform the attack on an FPGA-based convolutional neural network accelerator and we manage to recover the input image from the collected power traces without knowing the detailed parameters in the neural network. For the MNIST dataset, our power side-channel attack is able to achieve up to 89% recognition accuracy.

Key facts. Cited by 27 papers in the map (data through 14 Oct 2026). First posted on arXiv 5 Mar 2018.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2021: 10; 2022: 4; 2023: 5; 2024: 3; 2025: 4; 2026: 1.

27 papers.

Sort:

By area:

  1. 2024Operationalizing a Threat Model for Red-Teaming Large Language Models (LLMs)cited by 46
  2. 2021Robust Machine Learning Systems: Challenges, Current Trends, Perspectives, and the Road Aheadcited by 41
  3. 2023Privacy Side Channels in Machine Learning Systemscited by 40
  4. 2024What Was Your Prompt? A Remote Keylogging Attack on AI Assistantscited by 26
  5. 2021Pervasive AI for IoT applications: A Survey on Resource-efficient Distributed Artificial Intelligencecited by 18
  6. 2021Towards Energy-Efficient and Secure Edge AI: A Cross-Layer Frameworkcited by 17
  7. 2021Leaky Nets: Recovering Embedded Neural Network Models and Inputs through Simple Power and Timing Side-Channels -- Attacks and Defensescited by 11
  8. 2021Can one hear the shape of a neural network?: Snooping the GPU via Magnetic Side Channelcited by 8
  9. 2023FedRight: An Effective Model Copyright Protection for Federated Learningcited by 8
  10. 2021Knowledge Cross-Distillation for Membership Privacycited by 6
  11. 2022Special Session: Towards an Agile Design Methodology for Efficient, Reliable, and Secure ML Systemscited by 6
  12. 2022Careful What You Wish For: on the Extraction of Adversarially Trained Modelscited by 6
15 more
  1. 2021Physical Side-Channel Attacks on Embedded Neural Networks: A Surveycited by 5
  2. 2021A Review of Confidentiality Threats Against Embedded Neural Network Modelscited by 4
  3. 2021Confidential Machine Learning Computation in Untrusted Environments: A Systems Security Perspectivecited by 4
  4. 2023Mercury: An Automated Remote Side-channel Attack to Nvidia Deep Learning Acceleratorcited by 4
  5. 2021Model Extraction and Adversarial Attacks on Neural Networks using Switching Power Informationcited by 2
  6. 2023PowerGAN: A Machine Learning Approach for Power Side-Channel Attack on Compute-in-Memory Acceleratorscited by 2
  7. 2025THEMIS: Towards Practical Intellectual Property Protection for Post-Deployment On-Device Deep Learning Modelscited by 2
  8. 2025Stealing AI Model Weights Through Covert Communication Channelscited by 2
  9. 2022DPSNN: A Differentially Private Spiking Neural Network with Temporal Enhanced Poolingcited by 1
  10. 2025TAO: Tolerance-Aware Optimistic Verification for Floating-Point Neural Networkscited by 1
  11. 2022On the Evaluation of User Privacy in Deep Neural Networks using Timing Side Channelcited by 0
  12. 2023Side Channel-Assisted Inference Leakage from Machine Learning-based ECG Classificationcited by 0
  13. 2024Stealing the Invisible: Unveiling Pre-Trained CNN Models through Adversarial Examples and Timing Side-Channelscited by 0
  14. 2025DivQAT: Enhancing Robustness of Quantized Convolutional Neural Networks against Model Extraction Attackscited by 0
  15. 2026Guess My Weight: Profiled Side-Channel Recovery of Floating-Point Neural-Network Weightscited by 0