Work

Vulnerability analysis of transformer-based optical character recognition to adversarial attacks

Year: 2026

Type: article

Abstract: We present a novel framework to assess the resilience of state-of-the-art transformer-based optical character recognition (TrOCR) models. In this way, we develop new untargeted and targeted attack algorithms. On a benchmark handwriting dataset, we show that adversarial perturbations to the input reveal dramatic vulnerabilities without being noticeable to the eye. This article is part of the theme issue 'Safe, secure and robust AI for safety-critical systems'.

Source: Philosophical Transactions of the Royal Society A Mathematical Physical and Engineering Sciences

DOI: 10.1098/rsta.2025.0127

Authors: Lucas Beerens, Desmond J. Higham

Institution: University of Edinburgh

Language: English

Topic: Adversarial Robustness in Machine Learning

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: hybrid

Funders: Leverhulme Trust, Engineering and Physical Sciences Research Council

Awards: Inference, COmputation and Numerics for Insights into Cities (ICONIC), Mathematics of Adversarial Attacks, EPSRC Centre for Doctoral Training in Mathematical Modelling, Analysis and Computation (MAC-MIGS)