Skip to content
DraconiticResearch
Search Sources API
203,234 studies31 sourcesUpdated 52 min ago
Peer-reviewedCC BY 4.0The European Physical Journal C

Shapes are not enough: CONSERVAttack and its use for finding vulnerabilities and uncertainties in machine learning applications

  • Philip Bechtle1
  • Lucie Flek2
  • Philipp Alexander Jung3
  • Akbar Karimi2
  • Timo Saala1
  • Alexander Schmidt3
  • Matthias Schott1
  • Philipp Soldin3
  • Christopher Wiebusch3
  • Ulrich Willemsen3
  1. 1University of Bonn
  2. 2Bonn-Aachen Institute of Technology
  3. 3RWTH Aachen University
Open at sourcePDFDOICite

Abstract

In High Energy Physics, as in many other fields of science, the application of machine learning techniques has been crucial in advancing our understanding of fundamental phenomena. Increasingly, deep learning models are applied to analyze both simulated and experimental data. In most experiments, a rigorous regime of testing for physically motivated systematic uncertainties is in place. The numerical evaluation of these tests for differences between the data on the one side and simulations on the other side quantifies the effect of potential sources of mismodelling on the machine learning output. In addition, thorough comparisons of marginal distributions and (linear) feature correlations between data and simulation in “control regions” are applied. However, the guidance by physical motivation, and the need to constrain comparisons to specific regions, does not guarantee that all possible sources of deviations have been accounted for. We therefore propose a new adversarial attack – the CONSERVAttack – designed to exploit the remaining space of hypothetical deviations between simulation and data after the above mentioned tests. The resulting adversarial perturbations are consistent within the uncertainty bounds — evading standard validation checks — while successfully fooling the underlying model. We further propose strategies to mitigate such vulnerabilities and argue that robustness to adversarial effects must be considered when interpreting results from deep learning in particle physics.

Related

  1. Higgs signal strength estimation with machine learning under systematic uncertainties
    Minxuan He, Claudius Krause, Daohan Wang
    SciPostPeer-reviewedCC BY 4.0Full textGeneral physicsSciPostCrossref
  2. Learning Astrophysical Uncertainties in Dark Matter Direct Detection with Simulation-Based Inference
    Felix Kahlhoefer, Niklas Reus
    arXivPreprintCC BY 4.0Full textHigh-energy & particle physicsarXiv
  3. Bridging Control, Inference, Transport, and Thermodynamics: From Theory to Applications in Learning
    Emmy Blumenthal, Nikolas Claussen, Benjamin Eysenbach +4
    arXivPreprintCC BY 4.0Full textBiological & medical physicsCondensed matterarXiv
  4. Regret Lower Bound for Online Learning of Quantum States with Logarithmic Loss, with Applications to Quantum State Tomography and Confidence Sequences for Matrix Means
    Yen-Huan Li, Wei-Fu Tseng
    arXivPreprintarXiv licenceFull textStatisticsQuantum physics & quantum informationarXiv
  5. Machine Learning for High-Entropy Catalysts: Methods and Applications
    Hao Chen, Zongrui Pei, Xianglin Liu
    arXivPreprintCC BY 4.0Full textCondensed matterarXiv

Cite

BibTeXRIS
BibTeX
@article{bechtle2026shapes,
  title = {{Shapes are not enough: CONSERVAttack and its use for finding vulnerabilities and uncertainties in machine learning applications}},
  author = {Philip Bechtle and Lucie Flek and Philipp Alexander Jung and Akbar Karimi and Timo Saala and Alexander Schmidt and Matthias Schott and Philipp Soldin and Christopher Wiebusch and Ulrich Willemsen},
  journal = {The European Physical Journal C},
  year = {2026},
  month = oct,
  doi = {10.1140/epjc/s10052-026-16351-2},
  url = {https://link.springer.com/article/10.1140/epjc/s10052-026-16351-2},
  note = {CC BY 4.0},
}
RIS
TY  - JOUR
TI  - Shapes are not enough: CONSERVAttack and its use for finding vulnerabilities and uncertainties in machine learning applications
AU  - Philip Bechtle
AU  - Lucie Flek
AU  - Philipp Alexander Jung
AU  - Akbar Karimi
AU  - Timo Saala
AU  - Alexander Schmidt
AU  - Matthias Schott
AU  - Philipp Soldin
AU  - Christopher Wiebusch
AU  - Ulrich Willemsen
PY  - 2026
DA  - 2026/10/16
JO  - The European Physical Journal C
DO  - 10.1140/epjc/s10052-026-16351-2
UR  - https://link.springer.com/article/10.1140/epjc/s10052-026-16351-2
AB  - In High Energy Physics, as in many other fields of science, the application of machine learning techniques has been crucial in advancing our understanding of fundamental phenomena. Increasingly, deep learning models are applied to analyze both simulated and experimental data. In most experiments, a rigorous regime of testing for physically motivated systematic uncertainties is in place. The numerical evaluation of these tests for differences between the data on the one side and simulations on the other side quantifies the effect of potential sources of mismodelling on the machine learning output. In addition, thorough comparisons of marginal distributions and (linear) feature correlations between data and simulation in “control regions” are applied. However, the guidance by physical motivation, and the need to constrain comparisons to specific regions, does not guarantee that all possible sources of deviations have been accounted for. We therefore propose a new adversarial attack – the CONSERVAttack – designed to exploit the remaining space of hypothetical deviations between simulation and data after the above mentioned tests. The resulting adversarial perturbations are consistent within the uncertainty bounds — evading standard validation checks — while successfully fooling the underlying model. We further propose strategies to mitigate such vulnerabilities and argue that robustness to adversarial effects must be considered when interpreting results from deep learning in particle physics.
SN  - 1434-6044
SN  - 1434-6052
ER  -