Research map / Adversarial robustness and security / Attacks on text and graphs

Does Knowing More Make You Easier to Trick? Adversarial Robustness of Multi-Target Regression

AAAI 2025 · Adversarial robustness and security › Attacks on text and graphs · approach: Defenses against universal perturbations · 0 citations (OpenAlex)

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes a student research project studying adversarial robustness of linear multi-target regressors versus single-target regressors under deployment-time attractive attacks, aiming for controlled experiments and formal bounds.

Problem

Whether increasing the dimensionality of the target (multi-target regression) makes a model more or less susceptible to test-time adversarial attacks, motivated by understanding what large deep learning and GenAI models learn.

Approach

Uses the tractable setting of linear multi-target regressors (reduced-rank regression, FICYREG, SVS) under deployment-time attractive attacks with closed-form optimal perturbations, and links rank, row-wise L2 norm, and spectral norm to adversarial vulnerability.

Key results

• No results are reported; the paper is a proposal describing planned experiments and analysis.

Limitations

• No experimental results; only a proposed study
• Restricted to smaller-scale linear multi-output regression, with extrapolation to deep models left open

Experimental setup

Metrics: distance of attacked output from target

Baselines: single-target linear regressors

Builds on: reduced-rank regression FICYREG simultaneous variable selection Alfeld et al. adversarial attacks on linear predictors

Closest papers by meaning

• A Direct Approach to Robust Deep Learning Using Adversarial Networks (ICLR 2019 · similarity 0.84)
• Holistic Adversarial Robustness of Deep Learning Models (AAAI 2023 · similarity 0.83)
• On the Convergence and Robustness of Adversarial Training (ICML 2019 · similarity 0.83)
• Robust Models are less Over-Confident (NeurIPS 2022 · similarity 0.83)
• What It Thinks Is Important Is Important: Robustness Transfers Through Input Gradients (CVPR 2020 · similarity 0.83)
• Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.83)