ChatGPT as Preprocessing Agents: A Case Study on Cryptographic Side-Channel Analysis

  • Zhen Li
  • , Anjiang Liu
  • , An Wang*
  • , Wei Jia Wang
  • *Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Side-channel analysis techniques extract cryptographic keys by analyzing physical or electrical characteristics generated during the encryption process. When performing correlation power analysis and simple power analysis on raw traces, challenges such as noise interference necessitate effective trace preprocessing–a task that traditionally relies on domain expertise, specialized tools, and extensive experience. Meanwhile, ChatGPT has gained widespread attention for its intelligent interaction capabilities and effectiveness in assisting users with task-specific operations. However, its potential for trace preprocessing remains underexplored and calls for systematic investigation. In this paper, we propose three expert-strategy prompt templates to explore and assess ChatGPT’s capabilities in trace preprocessing. We validate the effectiveness of ChatGPT in performing six categories of trace preprocessing methods through expert-strategy prompts at varying abstraction levels. Furthermore, we evaluate the impact of ChatGPT-assisted preprocessing on traces across various platforms and cryptographic algorithms, analyzing its influence on the overall performance of side-channel analysis.

Original languageEnglish
Title of host publicationApplied Cryptography and Network Security Workshops - ACNS 2025 Satellite Workshops
Subtitle of host publicationAIHWS, AIoTS, QSHC, SCI, PrivCrypt, SPIQE, SiMLA, and CIMSS 2025, Revised Selected Papers
EditorsMark Manulis
PublisherSpringer Science and Business Media Deutschland GmbH
Pages193-210
Number of pages18
ISBN (Print)9783032018052
DOIs
Publication statusPublished - 2026
Externally publishedYes
EventSatellite Workshops held in parallel with the 23rd International Conference on Applied Cryptography and Network Security, ACNS 2025 - Munich, Germany
Duration: 23 Jun 202526 Jun 2025

Publication series

NameLecture Notes in Computer Science
Volume15654 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

ConferenceSatellite Workshops held in parallel with the 23rd International Conference on Applied Cryptography and Network Security, ACNS 2025
Country/TerritoryGermany
CityMunich
Period23/06/2526/06/25

Keywords

  • Cryptography
  • Large Language Model
  • Machine Learning
  • Side-Channel Analysis

Fingerprint

Dive into the research topics of 'ChatGPT as Preprocessing Agents: A Case Study on Cryptographic Side-Channel Analysis'. Together they form a unique fingerprint.

Cite this