Plausible Adversarial Attacks on Direct Parameter Inference Models in Astrophysics

2022/11/27 by Horowitz, Benjamin, Melchior, Peter
#Cosmology and Nongalactic Astrophysics (astro-ph.CO) #FOS: Physical sciences

paper · doi:10.48550/arxiv.2211.14788

Abstract

In this abstract we explore the possibility of introducing biases in physical parameter inference models from adversarial-type attacks. In particular, we inject small amplitude systematics into inputs to a mixture density networks tasked with inferring cosmological parameters from observed data. The systematics are constructed analogously to white-box adversarial attacks. We find that the analysis network can be tricked into spurious detection of new physics in cases where standard cosmological estimators would be insensitive. This calls into question the robustness of such networks and their utility for reliably detecting new physics.

Related

• Enforcing Fundamental Relations via Adversarial Attacks on Input Parameter Correlations
• Quantification of Unknown Unknowns in Astronomy and Physics
• Inferring Cosmological Parameters with Evidential Physics-Informed Neural Networks
• Detection is truncation: studying source populations with truncated marginal neural ratio estimation
• Cosmological Field Emulation and Parameter Inference with Diffusion Models