Forgery quality and its implications for behavioral biometric security.

Area of Science:

• Computer Science
• Cybersecurity
• Human-Computer Interaction

Background:

• Biometric security is crucial for user authentication and cryptographic key generation.
• Current evaluation methods often overlook sophisticated threat models.
• The assumption of minimally motivated forgers is dangerously optimistic.

Purpose of the Study:

• To develop novel evaluation methodologies for behavioral biometrics.
• To address under-explored threat models in biometric security.
• To assess the security implications of skilled human and generative forgers.

Main Methods:

• Conducted experiments using handwriting as a case study.
• Trained users to act as forgers to assess threat levels.
• Developed a generative attack model based on concatenative synthesis.
• Compared the effectiveness of human forgers and generative attacks.

Main Results:

• Identified users with significant forgery skills, trainable to increase threat.
• Demonstrated that some users are easy targets for forgers.
• Showed that human judgment of handwriting authenticity is unreliable.
• Generative attacks matched or exceeded the effectiveness of skilled human forgers.

Conclusions:

• Existing assumptions about forger motivation in behavioral biometrics are flawed.
• Human intuition is insufficient for authenticating handwriting.
• Advanced generative attack models are necessary for realistic security assessments.
• Developed methodologies provide a rapid indication of system security against sophisticated threats.