Description

Instructor:
Walter Scheirer
Office: 321C Stinson-Remick
Office Hours: 2:00-4:00, Mon, Weds
Email: wscheire@nd.edu
GTA:
Sreya Banerjee
Office: 212 Cushing
Office Hours: 3:00-5:00, Fri
Email: sbanerj2@nd.edu
Lecture Time and Place:
12:50-1:40, Mon, Weds, Fri,
DeBartolo Hall 116

From border crossings to social media, the ability of machines to determine human identity from images has matured to the point where end-users are now interacting with biometric algorithms on a daily basis. However, large-scale deployments in real-world settings bring with them new challenges to a system’s accuracy, security and privacy - thus biometrics remains a lively area of research. This course will introduce students to the fundamentals of biometric system development and evaluation, with an emphasis on the three primary visual modalities: face, fingerprint, and iris. Building from the seminal recognition algorithms, the course will cover the specialized feature descriptors and classification methods most commonly found in biometric application development, as well as the state-of-the-art deep learning architectures that are changing the way we think about visual recognition. Critical for social acceptance, the protection of biometric data is a growing concern. The final part of this course will describe the unique blend of cryptography and pattern recognition that is being developed to revoke and reissue protected biometric data like passwords or PINs.

Objectives

By the end of the course you will be able to:

  • Describe the principles of the three core biometric modalities (face, fingerprint and iris), and know how to deploy them in authentication scenarios;
  • Organize and conduct biometric data collections, and apply biometric databases in system evaluation;
  • Calculate distributions of within- and between-class matching scores, and calculate various error estimates based on these distributions;
  • Identify the privacy and security concerns surrounding biometric systems, and know how to address them in such a way that balances both;
  • Recognize differences between algorithm design and systems engineering in biometrics;
  • Deploy statistical methods in biometric system evaluation;
  • Itemize the most up-to-date examples of real biometric applications in human authentication.

Lectures (Synopsis)

The Basics of Biometrics (4 Lectures). Overview of field and applications. Development of biometric authentication. Basic terms, biometric data, biometric characteristics, biometric features, biometric templates and references. Expected properties of biometric identifiers. Basics in biometric errors estimation. Enrollment, verification and identification.

Face Recognition (12 Lectures). Introduction to the face processing pipeline: acquisition, face detection, alignment, feature extraction, matching. Classic subspace methods. Hand-tuned feature descriptors. Deep learning architectures for face representation learning. Distance, similarity and learning-based matching. Face recognition in video. Describable visual attributes. Face pair matching, verification, and identification. Data sets for evaluation. Face image quality. Considerations for social media, mobile authentication, surveillance and other real-world applications.

Fingerprint Recognition (6 Lectures). Fingerprint capture, sensor types, latent fingerprints. Fingerprint image preprocessing, segmentation, binary and skeletal images. Fingerprint singularities, detection of loops, deltas, whirls and cores, using singularities in fingerprints classification. Galton's details, base and complex minutiae, detection of minutiae. Fingerprint recognition, minutiae- and correlation-based methods. Fingerprints in forensics and biometrics, similarities and differences.

Iris Recognition (6 Lectures). Eye and iris morphogenesis, genetic penetrance. Principles of iris image capture, iris sensors. Iris image preprocessing, segmentation, formatting and filtering. Daugman’s method, iris code, statistical properties of the iris code. Other iris coding methods, wavelet analysis.

Multi-Biometric Fusion (3 Lectures). Levels of fusion: sensor, feature, rank, decision. Score normalization and fusion rules. Quality-based fusion and failure prediction.

Biometric System Security (1 Lecture). Secure transfer of biometric data. Secure storage, use of smart cards, principles of match-off-card and match-on-card techniques. Biometrics in the cloud. Points of attack. Privacy models.

Spoofing (2 Lectures) Static and dynamic liveness features. What we want to detect (subversive actions) vs. what we can detect (suspicious actions). Liveness detection in biometrics. Selected liveness detection techniques, frequency analysis for paper printouts detection, pupil dynamics and blood pulse analyses for detection of sophisticated eye and finger spoofing trials.

Template Protection (5 Lectures) Overview of principles from cryptography that help us secure fuzzy data. Template protection strategies: feature protection, key-binding, key-generating, hybrids. Overview of fuzzy vaults, fuzzy commitment, fuzzy extractors and revocable biotokens. Biocryptographic infrastructures for secure template management.

Invited Talks (2 Lectures) Identical Twins and Face Recognition (Prof. Bowyer) and Biometric Data Sets and Evaluation Research at Notre Dame (Prof. Flynn).

Prerequisites

Courses in statistics (estimation and hypotheses testing), image processing (1D and 2D filtering), and basic principles of symmetric and asymmetric cryptography will help in faster adoption of the discussed topics, but are not necessarily required.

Textbook

There are NO required textbooks for this course!

Does this means there are no written materials relevant to this course? Of course not! It means up-to-date information about biometrics is now available on the web. This material is both cheaper and easier to access than traditional textbooks. The resources page of this website contains links to texts that supplement the content covered in the lectures. Interested students may find these texts to be a good starting point for original research down the road.

Grading

Here are the formally graded elements of the course and associated weighting:

Activity Weight Points
Class Participation 10 % 100
Homework Assignments (4) 70 % 175 x 4
Final Exam 20 % 200

Class participation includes much more than just attendance. It includes answering questions in class when called upon, and most importantly asking questions and initiating discussions in class. The final exam will cover material from the entire course, with an emphasis on material not covered by the programming assignments. There will be four homework assignments that will require programming and data analysis, with code and a written report as the deliverables.

Important Dates

Homework #1 (Face Detection)Released: Monday, September 14th; Due: Monday, September 28th
Homework #2 (Face Recognition) Released: Friday, October 2nd; Due: Friday, October 16th
Homework #3 (Iris Recognition)Released: Friday, October 30th; Due: Friday, November 13th
Homework #4 (Fingerprint Spoof Detection) Released: Monday, November 23rd; Due: Monday, December 7th
Final Exam Thursday, December 17th, 8-10am

Policies

Honor Code. All work that you submit must be your own. You may discuss assignments with other students or refer to books or websites as long as you cite your sources. You may not write solutions or code with other students or anyone else, nor may you copy solutions from any source. In certain cases, permission will be given to reuse publically available code. Assignments will indicate when this is possible. The university's honor code site is a good resource for understanding what is and is not permissible.

Late and Makeup Policy. In the case of a serious illness or other excused absence, as defined by university policies, coursework submissions will be accepted late by the same number of days as the excused absence. A make-up exam will only be given under extraordinary circumstances. In such cases, a student must consult with the instructor as soon as possible, preferably before the start of the exam.

Students with Disabilities. Any student who has a documented disability and is registered with Disability Services should speak with the professor as soon as possible regarding accommodations. Students who are not registered should contact the Office of Disability Services.