Biometric Recognition: Modalities and Authentication

Biometric Recognition

• Biometrics is the use of physical or behavioral human traits for automatic identification. It's unique, unlike something you know (password) or have (key).

Biometric Recognition Types

• Positive Recognition: Matching a sample to a known, registered individual.
• Negative Recognition: Determining a sample does not match any registered subject.

Biometric Modality

• A single physical or behavioral trait used in biometric recognition (e.g., face, fingerprint, iris).

Probe and Gallery

• Probe: The sample presented to the biometric system for identification.
• Gallery: A collection of previously enrolled templates.

Authentication Types

• Pair Matching: Checks if two samples match.
• Verification: Confirms a sample matches a specific registered individual (1:1 match).
• Identification: Determines if the individual is in the system, matching against multiple templates (1:n match).
• Negative Authentication: Asserts an individual is not in a specified group or subject list.
• Deduplication: Checks if a subject is already enrolled to avoid redundant registrations

Enrollment Pipeline (New User)

• Acquisition: Raw data collection from a sensor.
• Feature Extraction: Transforming raw data into a usable representation.
• Template Generation: Creating a unique representation (a template) for the user, stored in the system.

Recognition Pipeline (Decision Making)

• Acquisition: Collecting the biometric sample.
• Feature Extraction: Transforming the raw data into features.
• Matching: Comparing the extracted features to stored templates.
• Labeling: Determining a match or non-match based on the comparison scores and set thresholds.

Key Concepts

• Features: Mathematical representations of biometric data (x=F(I, φF),).
• Templates: Unique representations of individuals stored in the system (y = label, (x1,y1),…) for a given biometric modality.
• Matching: Comparing an input sample to stored templates using a similarity score (sy).
• Labeling: Assigning a match/non-match label based on the similarity scores (C = L(S, φL)).

Enrollment - User Perspective

• Multiple measurements: Collection of multiple samples for quality control and best sample selection.

Enrollment - Security Engineer's Perspective

• Metadata: Additional data associated with a user's record (important for security and context).
• Multiple Templates: Potentially multiple templates stored for a user (e.g., from different sensor readings or time points).
• Operator Supervision: Oversight of the enrollment process.
• Template Storage: Secure storage methods, including encryption and template protection.

Authentication - User Perspective

• How templates are retrieved: How templates are accessed dependent on the user's access type and device location .
• Matching Threshold: Criteria for determining a successful match or authentication.
• Attempts permitted: The number of authentication attempts allowed (security measure).
• Speed vs. Security: Balancing faster processing speeds and security robustness.
• Auditing: Keeping a detailed record of authentication attempts and events (for accountability).

Authentication - Security Engineer's Perspective

• Template Retrieval: How the appropriate templates are accessed for matching (on-device storage, cloud servers, etc.).
• Operational Thresholds: Rules and procedures the systems use for matching attempts to confirm identities.
• Auditing: A systematic recording of each authentication attempt for tracking security incidents and regulatory compliance.

Explore biometric recognition, which uses unique human traits for automatic identification. Learn about positive and negative recognition, biometric modalities like face or fingerprint, and concepts like probe and gallery. Understand pair matching, verification, identification, and negative authentication.