Paper 2020/988

120.147 Efficient Electromagnetic Side Channel Analysis by Probe Positioning using Multi-Layer Perceptron

Anupam Golder, Baogeng Ma, Debayan Das, Josef Danial, Shreyas Sen, and Arijit Raychowdhury

Abstract

In this work, we investigate a practical consideration for Electromagnetic (EM) side-channel analysis, namely, positioning EM probe at the best location for an efficient attack, requiring fewer traces to reveal the secret key of cryptographic engines. We present Multi-Layer Perceptron (MLP) based probe positioning and EM analysis method, defining it as a classification problem by dividing the chip surface scanned by the EM probe into virtual grids, and identifying each grid location by a class label. The MLP, trained to identify the location given a single EM trace, achieves $99.55\%$ accuracy on average for traces captured during different acquisition campaigns.

Note: This work has been presented as a poster at 57th Design Automation Conference, July 20-24, 2020. (http://www2.dac.com/events/eventdetails.aspx?id=295-120). This is the full paper version of the poster.

Metadata
Available format(s)
PDF
Category
Applications
Publication info
Published elsewhere. 57th DAC Work-In-Progress Session, 2020
Keywords
EM Probe PositioningSide-Channel AnalysisMulti-Layer PerceptronCorrelation Analysis
Contact author(s)
anupamgolder @ gatech edu
History
2020-08-18: received
Short URL
https://ia.cr/2020/988
License
Creative Commons Attribution
CC BY

BibTeX

@misc{cryptoeprint:2020/988,
      author = {Anupam Golder and Baogeng Ma and Debayan Das and Josef Danial and Shreyas Sen and Arijit Raychowdhury},
      title = {120.147 Efficient Electromagnetic Side Channel Analysis by Probe Positioning using Multi-Layer Perceptron},
      howpublished = {Cryptology ePrint Archive, Paper 2020/988},
      year = {2020},
      note = {\url{https://eprint.iacr.org/2020/988}},
      url = {https://eprint.iacr.org/2020/988}
}
Note: In order to protect the privacy of readers, eprint.iacr.org does not use cookies or embedded third party content.