Adversarial Attack Evaluation in Offline Handwritten Signature Verification Systems

Key Points

• The aim is to assess how adversarial attacks affect the effectiveness of handwritten signature verification systems.
• Conducted evaluations of various signature verification systems under adversarial attack scenarios.
• Analyzed the resilience of these systems against specific attack configurations using machine learning techniques.
• Implemented performance metrics to quantify the impact of attacks on verification accuracy.
• Found significant vulnerabilities in existing signature verification systems against adversarial attacks.
• Demonstrated that certain systems were more resilient than others, providing insights on improving security measures.
• Reported variations in accuracy and reliability based on the type of adversarial attacks employed.

Abstract