search

biometricITC / cPP-toolboxes

This repository maintains the toolbox definitions for evaluations according to the biometric cPP
2 stars 1 forks source link

2D face image toolbox update #29

Closed n-kai closed 4 years ago

n-kai commented 4 years ago

I largely update and simplify the 2D face image toolbox. I will update 3D face and vein if this PR is approved. I also made comments on PP-Module, SD, PP-Conf and Toolbox overview during this task to keep consistency between toolbox and those other document (but most of my comments are editorial fixes).

woodbe commented 4 years ago

So a question I have here is if we need to (at least in this particular case) have 6 total use case documents or if we can just have 3 with 2 variations in each.

In my eye toolbox, the Variations section is used not to describe things the evaluator can tweak in looking for the best artefact, but different artefacts to be generated by the same procedures with different tools/media. So as a simple example (not relevant here, but simple), the main test could be to take the picture and print it on regular office paper (in color using an inkjet). The variation would be the same thing but using photo paper.

So if you were going to create 3 artefacts for the normal paper, you would need to do another 3 for the photo paper.

The point for this was to minimize the total number of documents that had to be maintained, especially when you were doing the same thing over an over (that is also why some ended up multiple attack potentials if there were possible differences).

So I think it would be a good idea to do that were we can, just to minimize maintenance over the long term. I can see places where there would not be both levels and so there would be attacks where there may not be any variations (or they would be things like the ISO levels or something that don't have anything to do with the normal vs high), so I'm not precluding this idea, only thinking about how to streamline our docs for what we are proposing right now.

n-kai commented 4 years ago

@woodbe I agree. I also realized during making 6 use cases that they are very similar. I will look at eye toolbox and reduce the number of use cases to 3.

n-kai commented 4 years ago

I updated 2D face toolbox to reduce the number of use case using Variations. Current Toolbox overview's definition of Variations should also be changed as proposed by @woodbe so I create PR #32 to correct this

woodbe commented 4 years ago

@n-kai do we need to show both the exploitation and identification phases of the attack potential?

Overall this looks pretty good, but I will make a few edits based on that and question (as I will also then update the eye to match).

n-kai commented 4 years ago

@woodbe do we need to show both the exploitation and identification phases of the attack potential?

Yes because we have to follow the attack potential table defined in Table 3 of 9.1.3. Calculation of attack potential in BIOSD. https://github.com/biometricITC/cPP-biometrics/blob/master/4_Methodology/BS_SD.adoc#attackpotentialcalc

This table is also used in ISO/IEC 19989 so it's better to follow this table