29
Pattern Recognition and Applications Lab Università degli Studi di Cagliari, Italia Dipartimento di Ingegneria Elettrica ed Elettronica Biometric technologies and behavioral security Gian Luca Marcialis [email protected] https://people.unica.it/gianlucamarcialis/ M.Sc. Degree In Computer Engineering, CyberSecurity and Artificial Intelligence

Biometric technologies and behavioral security

  • Upload
    others

  • View
    7

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Biometric technologies and behavioral security

Pattern Recognitionand Applications Lab

Università degli Studidi Cagliari, Italia

Dipartimento diIngegneria Elettrica

ed Elettronica

Biometric technologiesand behavioral security

Gian Luca Marcialis

[email protected]

https://people.unica.it/gianlucamarcialis/

M.Sc. Degree In Computer Engineering, CyberSecurity and Artificial Intelligence

Page 2: Biometric technologies and behavioral security

http://pralab.diee.unica.it

Biometrics @ PRA Lab

http://pralab.diee.unica.it/en/Biometrics

2

Page 3: Biometric technologies and behavioral security

http://pralab.diee.unica.it

Overview of the course

• Biometric technologies– Definition

– Fingerprints

– Faces

– Multiple biometrics

– Other biometrics

• Security and biometrics– Vulnerabilities of biometric systems

– Presentation Attacks Detection

– Design of countermeasures against spoofing attacks

• Behavioural security– Crowd analysis

– Anomalous behaviour

– Human-in-the-loop

3

Page 4: Biometric technologies and behavioral security

http://pralab.diee.unica.it

Course organization

• Lectures and laboratory exercises– Monday, 9am, LIDIA room (2 hours)

– Thursday, 8am, LIDIA room (3 hours)

• Balance between lectures and exercises

4

Lectures58%

Exercises42%

Page 5: Biometric technologies and behavioral security

http://pralab.diee.unica.it

What will you learn during lectures?

• Basics of biometric systems design and evaluation

• Main modules of a biometric system

• How pattern recognition and machine learning concepts are applied to biometric recognition

• Design of presentation attacks detectors as a basic protection«infrastructures» for biometric systems

• How people behavior can be «detected» for mass-oriented video-surveillance systems

5

Page 6: Biometric technologies and behavioral security

http://pralab.diee.unica.it

What will you learn to do during the laboratory exercises?

• To design and test biometric recognition systems

• Using advanced programming languages and tools for biometricapplications– Matlab/Python/C

– Image processing

– Pattern recognition

– Deep learning

• Exploring pros and cons of video-based analysis

6

Page 7: Biometric technologies and behavioral security

http://pralab.diee.unica.it

How to gain the exam score?

• By following the course

– Last 5 hours of laboratory exercises will be a true biometric system design test

– This can be carried out alone or by group of two/three people maximum

• By the M.Sc. Thesis

– In this case, it is not necessary to develop anything more than a good topic, under my supervision

• By an oral verification on the whole course program

– 60/90 minutes of interview with me and my co-workers

– Cross your fingers and good luck!

7

Page 8: Biometric technologies and behavioral security

http://pralab.diee.unica.it

How to study

• Follow the course

• Do questions– During lecture/lab exercises

– During the break

– By e-mail

• Download the course slides– https://people.unica.it/gianlucamarcialis/

• Integrate your notes and book information (no wikipedia)

• Do programming exercises at home

8

Page 9: Biometric technologies and behavioral security

http://pralab.diee.unica.it

Bibliography

• A. Jain et al., Handbook of Biometrics, Springer

• B. Bhanu and A. Kumar, Deep learning in biometrics, Springer

• K. Saeed, New direction in behavioural biometrics, CRC Press

• V. Murino et al., Group and crowd behavior for computer vision, Academic Press

• D. Maltoni et al., Handbook of fingerprint recognition, Springer

• H. Liu, Face Detection and Recognition on Mobile Devices, Elsevier

• M. Vatsa et al., Deep learning in biometrics, CRC Press

9

Page 10: Biometric technologies and behavioral security

http://pralab.diee.unica.it

Use-inspired research

Did you seesomething?

No, it’s allright!

10

Page 11: Biometric technologies and behavioral security

http://pralab.diee.unica.it

Vulnerability points

11

Page 12: Biometric technologies and behavioral security

http://pralab.diee.unica.it

Fingerprint anti-spoofing

Stand-alone fingerprintverification systemwith anti-spoofing

tailored for differentcapture devices

12

Page 13: Biometric technologies and behavioral security

http://pralab.diee.unica.it

«Genuine» spoofs from our lab

13

Page 14: Biometric technologies and behavioral security

http://pralab.diee.unica.it

What about… deep fakes?

14

Page 15: Biometric technologies and behavioral security

http://pralab.diee.unica.it

Behavioural security: starting point

15

Page 16: Biometric technologies and behavioral security

http://pralab.diee.unica.it

Cluster detection

16

Page 17: Biometric technologies and behavioral security

http://pralab.diee.unica.it

A recent project: the BullyBuster

17

Bari – Cagliari – Foggia – Napoli «Federico II»

Page 18: Biometric technologies and behavioral security

http://pralab.diee.unica.it

Let’s start:the biometric paradigm

18

Page 19: Biometric technologies and behavioral security

http://pralab.diee.unica.it

Definition

19

Something you are

2

Page 20: Biometric technologies and behavioral security

http://pralab.diee.unica.it

Something that (only) you (should) know

20

Page 21: Biometric technologies and behavioral security

http://pralab.diee.unica.it

Identity theft and cybersecurity

21

Page 22: Biometric technologies and behavioral security

http://pralab.diee.unica.it

Biometric systems and «security chain»

• Identity check is the first stage in current «security chain»

22

Page 23: Biometric technologies and behavioral security

http://pralab.diee.unica.it

Biometric properties

23

Page 24: Biometric technologies and behavioral security

http://pralab.diee.unica.it

Biometrics for personal recognition

• Biometric systems are pattern recognition-based algorithms

• They require that the system is «trained»

– By collecting a set of the target users «templates»

– By embedding the uniqueness measurements of users as system’s parameters

24

Biometricsystem

Who is he?Is he who isclaiming to

be?

Page 25: Biometric technologies and behavioral security

http://pralab.diee.unica.it

Biometric system’s rough view

25

Feature

extraction Identification?

Verification

module

Identification

module

yes

no

Score or

distance set

Score or

distance

Identification module

Classification

module

Fingerprints

Data Base

Matching

module

Verification module

Fingerprints

Data Base

Matching

module

Claimed Identity

Page 26: Biometric technologies and behavioral security

http://pralab.diee.unica.it

Personal recognition…

26

26

AirportSUSPECT

Comparison with

a lot of possible

targets

Suspect subjects database

Page 27: Biometric technologies and behavioral security

http://pralab.diee.unica.it

…and personal verification

DATABASE of Templates

MATCHERFEATURE

EXTRACTOR

I verified this is

YOUR fingerprint

ACCESS GRANTED

This is my fingerprint –please grant me access to

my home

Page 28: Biometric technologies and behavioral security

http://pralab.diee.unica.it

That’s all for today

• Pattern Recognition: what is it?– Pre-processing

– «Features» extraction

– Machine learning

– Classification

• Some recalls about statistics– Random variables

– Distributions

– Hypothesis verification tests

• Don’t miss the next exciting lecture!

28

Page 29: Biometric technologies and behavioral security

http://pralab.diee.unica.it

Thank you for listening!

29

Gian Luca Marcialis

Phone: +39 070 675 5893E-mail: [email protected]: http://pralab.diee.unica.it

Università degli Studi di CagliariDip. Ing. Elettrica ed Elettronica