AI – Help or Hindrance?

As artificial intelligence becomes ever more important to document authentication, new research warns that its strengths can also be its weaknesses. A study presented at the 2025 Optical & Digital Document Security™ conference in Warsaw shows how adversarial attacks can fool even the most accurate AI systems into verifying counterfeit documents as genuine.

AI’s double-edged role

Artificial intelligence (AI) is transforming document verification – from passport inspection to mobile banknote authentication – by enabling machines to detect what the human eye might miss. Researchers at the Institut für industrielle Informationstechnik (inIT) in Germany have shown, however, that these same systems can be deceived by adversarial attacks: tiny, precisely engineered image manipulations that trigger misclassifications in neural networks.