CAD for Assurance of Electronic Systems

Machine Learning Model Correlated Encoding Quantization Attack Flow

By: Nuo Xu (Lehigh University), Qi Liu (Lehigh University), Wujie Wen (Lehigh University)

Stage: ML Model Training


This is a quantized correlation encoding attack flow from an adversary perspective. The tool can encode users’ training data into the model parameters with high quality, even on the extremely quantized DNN models, without compromising model accuracy.


Input/Output Interface

  • Input: Target dataset, model structure, and quantization bit number
  • Output: Quantized training model with target data encoded


Python, TensorFlow, Theano

Toolset can be found at this link


Xu, Nuo; Liu, Qi; Liu, Tao; Liu, Zihao; Guo, Xiaochen; Wen, Wujie

Stealing your data from compressed machine learning models Proceedings Article

In: 2020 57th ACM/IEEE Design Automation Conference (DAC), pp. 1–6, IEEE 2020.

Abstract | Links | BibTeX