Research map / Adversarial robustness and security / Certified and empirical robustness

Endowing Visual Reprogramming with Adversarial Robustness

ICLR 2025 · Adversarial robustness and security › Certified and empirical robustness · approach: Adversarial training refinements

Read the paperCheck your idea against 108,967 papers

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

Studies adversarial robustness of visual reprogramming (VR), which repurposes pre-trained ImageNet classifiers for target tasks without training from scratch, and proposes a method to endow reprogrammed models with that robustness.

Problem

Previous visual reprogramming methods do not consider adversarial robustness of reprogrammed models against adversarial attacks, which could lead to unpredictable problems in deployment.

Approach

Addresses adversarial robustness specifically within the visual reprogramming framework; the abstract states the gap but the specific technique is not described in the provided text.

Closest papers by meaning

• Robustness Reprogramming for Representation Learning (ICLR 2025 · similarity 0.90)
• Sample-specific Masks for Visual Reprogramming-based Prompting (ICML 2024 · similarity 0.87)
• Understanding Model Reprogramming for CLIP via Decoupling Visual Prompts (ICML 2025 · similarity 0.87)
• Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.86)
• Attribute-based Visual Reprogramming for Vision-Language Models (ICLR 2025 · similarity 0.86)
• TARP-VP: Towards Evaluation of Transferred Adversarial Robustness and Privacy on Label Mapping Visual Prompting Models (NeurIPS 2024 · similarity 0.86)