analiza aiqbench
This research addresses critical safety gaps in computer vision by introducing a certified training method that provides formal guarantees against runtime perturbations like motion blur, outperforming standard adversarial approaches. The work is significant because it moves beyond empirical robustness to offer provable defenses for deploying AI models in high-stakes environments where failure due to minor visual distortions could be catastrophic.
Certified Training for Convolutional Perturbations
arXiv:2607.18195v1 Announce Type: cross Abstract: Vision models have been found to be susceptible to perturbations such as motion blur induced at runtime by a shaking camera. This impedes their deployment in critical applications since phenomena such as slightly blurred vision might lead to failures, for example an object detector missing objects. While methods such as data augmentation or Adversarial Training can improve empirical robustness, they lack formal safety guarantees, making it difficult to identify and mitigate hidden vulnerabilities. We introduce a novel Certified Training approach that leverages an efficient encoding of convolutional perturbations to train provably robust models. Our method significantly outperforms Adversarial Training, achieving, for example, over 80% robust accuracy against motion blur of reasonable intensity on CIFAR10 while maintaining comparable standard accuracy.
Abstract:Vision models have been found to be susceptible to perturbations such as motion blur induced at runtime by a shaking camera. This impedes their deployment in critical applications since phenomena such as slightly blurred vision might lead to failures, for example an object detector missing objects. While methods such as data augmentation or Adversarial Training can improve empirical robustness, they lack formal safety guarantees, making it difficult to identify and mitigate hidden vulnerabilities. We introduce a novel Certified Training approach that leverages an efficient encoding of convolutional perturbations to train provably robust models. Our method significantly outperforms Adversarial Training, achieving, for example, over 80% robust accuracy against motion blur of reasonable intensity on CIFAR10 while maintaining comparable standard accuracy.
Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
arXiv:2607.18195 [cs.CV]
(or arXiv:2607.18195v1 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2607.18195 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 219 cuvinte
Subiecte conexe
Neural Network Verification for Deep Joint Source-Channel Coding
arXiv · 10 oct. 2026
S-GBT: Smooth Growth Bound Tensor for Certified Robustness Against Word Substitution Attacks in NLP
arXiv · 12 iun. 2026
Anatomically Faithful but Temporally Diffuse: Auditing Attribution for Left-Ventricular Ejection-Fraction Estimation from Echocardiography
arXiv · 24 iul. 2026
Certifying Lower Bounds for Risk-Sensitive Reinforcement Learning under Adversarial State Perturbations
arXiv · 11 sept. 2026