Reproduce “Adversarial AI in Insurance: Pervasiveness and Resilience” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Elisa Luciano, Matteo Cattaneo, Ron Kenett

Researchers from the University of Torino and Reale Mutua Group examine the widespread threat of adversarial AI across the insurance value chain, from product design to claims processing, and propose resilience strategies that emphasize zero-shot and few-shot multi-labelling techniques to counter deceptive, often imperceptible, attacks.

Key insights

  • Adversarial AI attacks are pervasive, impacting every stage of the insurance value chain from product design and risk assessment to claims processing.
  • Adversarial examples are often imperceptible to humans and can lead AI systems to misclassify with high confidence, sometimes even appearing to improve system accuracy.
  • Zero-shot and few-shot multi-labelling techniques are under-recognized yet highly relevant for defending against novel, 'never-seen-before' adversarial attack patterns.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.