TY - GEN
T1 - ChatGPT as Preprocessing Agents
T2 - Satellite Workshops held in parallel with the 23rd International Conference on Applied Cryptography and Network Security, ACNS 2025
AU - Li, Zhen
AU - Liu, Anjiang
AU - Wang, An
AU - Wang, Wei Jia
N1 - Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2026.
PY - 2026
Y1 - 2026
N2 - 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.
AB - 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.
KW - Cryptography
KW - Large Language Model
KW - Machine Learning
KW - Side-Channel Analysis
UR - https://www.scopus.com/pages/publications/105021005936
U2 - 10.1007/978-3-032-01806-9_11
DO - 10.1007/978-3-032-01806-9_11
M3 - Conference contribution
AN - SCOPUS:105021005936
SN - 9783032018052
T3 - Lecture Notes in Computer Science
SP - 193
EP - 210
BT - Applied Cryptography and Network Security Workshops - ACNS 2025 Satellite Workshops
A2 - Manulis, Mark
PB - Springer Science and Business Media Deutschland GmbH
Y2 - 23 June 2025 through 26 June 2025
ER -