202604.16
International Journal on Document Analysis and Recognition (IJDAR)
PROFILE
Since offline handwritten signature identification systems are vulnerable to imperceptible tiny perturbations, this research studies adversarial attack methods for offline handwritten signature identification systems to test the security of signature‑verification systems. Meanwhile, to verify the effectiveness of attack methods, a defense method for offline handwritten signature identification is studied to resist adversarial attacks, balance attack and defense, and improve the security of signature systems.
Keywords: machine learning, pattern recognition, adversarial attack, adversarial defense
‑ National Third Prize for C Language in National University Computer Challenge ‑ Second Prize of Literature Review Competition