Biometrics and Machine Learning Group
Latest news
We are pleased to announce that Mateusz Trokielewicz defended (with honors) his doctoral dissertation entitled „Iris Recognition Methods Resistant to Biological Changes in the Eye” , supervised by prof. Czajka and prof. Pacut, on the 18th of July, 2019.
Iris scanner can distinguish dead eyeballs from living ones: MIT Technology Review reports on our recent developements in the field of presentation attack detection for cadaver irises.
We are pleased to announce that Mateusz Trokielewicz received the EAB European Biometrics Research Award 2016 for research on iris recognition reliability including template aging, influence of eye diseases and post-mortem recognition.
Is That Eyeball Dead or Alive? Adam Czajka discusses the prevention of iris sensors accepting the use of a high-resolution photo of an iris or, in a grislier scenario, an actual eyeball. For full article, please see IEEE Spectrum.
Biometrics (CSE 40537 / 60537)
back to Biometrics (CSE 40537/60537)
Quizzes
How to solve quizzes?
- Send your answers to aczajka@nd.edu by the due date indicated in each quiz. Please send your answers in plain text and don't use Word or PDF attachments (unless it is necessary).
- Provide your name or your netID in the email so that I can identify the author.
- Send your answer quickly. If it is incorrect, we will have some time to develop a correct answer until the deadline.
- We will discuss shortly the correct answers is class (after the deadline).
Quiz 8 (4 points)
Send your answers by Monday, April 30, 11:59 pm
- Do a Google search and find one example attack on a biometric system observed in the operational environment (either successful or not). It can be any type of attack: presentation, hill climbing, biometric data illicit re-use, etc. But it must be different than examples presented in class. Provide a link to the article / news you found.
- Propose a countermeasure that you would apply to detect this kind of attack. Justify briefly (a few sentences) your choice.
Quiz 7 (2 points)
Send your answers by Monday, April 23, 11:59 pm
- “Formants” in biometrics are:
- Quantities used to describe the properties of a human vocal track
- Characteristic quantities related only to the sound of vocal cords
- The title of Lucier’s famous composition
- Mel-frequency scale:
- Explains better than a linear scale the nature of how we (humans) hear the sounds
- There is only one and universal mel-scale (independent of the population)
- Since the name “mel” originates from “melody” this scale is used in all opera performances
Quiz 6 (2 points)
Send your answers by Monday, April 9, 11:59 pm
This time questions are in this PDF document.
Quiz No. 5 (2 points)
Send your answers by Monday, March 26, 11:59 pm
- Which statement about hand thermography is true:
- temporal stability of temperature-based hand features is high
- all parts of the hand heatmap are equally important in recognition
- it can be combined with hand geometry to increase the probability of detecting presentation attacks
- Different geometrical features of our hand have different variance. Which metric is better to calculate a comparison score in this case:
- Hamming
- Euclidean
- Mahalanobis
Quiz No. 4 (2 points)
Send your answers by Thursday, March 8, 11:59 pm
- Imagine that you are trying to reverse Daugman’s iris coding. That is, given the iris code C you want to find an iris image that generates this code. Choose a correct statement:
- reversing of Daugman’s coding is possible, and I will end up with a single and original iris image that generated the code C
- reversing of Daugman’s coding is possible, and I will be able to generate multiple images (iris and non-iris) that correspond to the same code C
- this operation (reversing of Daugman’s coding) is impossible
- Imagine that someone flipped the iris image along the horizontal axis to make it upside-down. Does the iris code calculated for such a flipped image match the iris code calculated for the upright image of the same eye?
- yes, it’s the same eye = the same features!
- yes, but only when a huge compensation for image rotation is applied
- no, the mapping of features will be different and cannot be corrected by shifting the iris codes
Quiz No. 3 (4 points)
Send your answers by Friday, March 2, 11:59 pm
- Which core difficult aspects of 2D face recognition become more tractable in 3D face recognition? Provide and describe at least three such aspects and justify your choice.
- Watch this short video. Starting from 2:55 it demonstrates a common problem of recognizing faces of monozygotic twins and triplets. Knowing that iPhone X implements 3D face scanning, what improvements to this algorithm would you suggest to increase recognition accuracy of twins? Tip: this survey paper may provide a few ideas.
Quiz No. 2 (2 points)
Send your answers by Monday, February 12, 11:59 pm
- Bifurcation is:
- a singular point
- a basic minutia
- a level-three feature
- Assume that Poincaré index is zero for all tested local areas of our fingerprint image. The Henry's class of this fingerprint is:
- whorl
- double loop
- plain arch
Quiz No. 1 (2 points)
Send your answers by Thursday, January 25, 11:59 pm
- Imagine that we decided to use biometric recognition to identify frequent travelers at Chicago O'Hare airport. Is DNA a good biometric mode in this application? Justify your answer.
- We have tested two biometric methods: A and B. The following error estimators were obtained: FNMRA=0.1, FMRA=0.02, FNMRB=0.01, FMRB=0.2. Which method would be better for applications requiring high security?