240 likes | 592 Views
Talk Outline. DefinitionsVirtual World and Security Preliminary WorkMethodologyDatabase collectionBiometric based avatar authenticationMultimodal fusion systemConclusions and Future Work. Virtual World and Security. Terrorists have been reported recruiting in virtual communities such as Second Life. Cybercrime, including identity theft, is rampant in virtual worlds populated by millions of avatars and operating multibillion dollar economies. Security experts have testified to the US Sen9457
E N D
1. Applying Biometric Principles to Avatar Recognition Marina L. Gavrilova
Roman V. Yampolskiy
Computer Science Dept, U. of Calgary,
Alberta, Canada
Comp. Science and Engineering, University of Louisville
Kentucky, USA
2. Talk Outline Definitions
Virtual World and Security
Preliminary Work
Methodology
Database collection
Biometric based avatar authentication
Multimodal fusion system
Conclusions and Future Work
3. Virtual World and Security Terrorists have been reported recruiting in virtual communities such as Second Life.
Cybercrime, including identity theft, is rampant in virtual worlds populated by millions of avatars and operating multibillion dollar economies.
Security experts have testified to the US Senate that defenses are lacking when it comes to the nation's Cyberinfrastructure.
International teams of hackers assisted hacking software agents have perpetrated numerous attacks against the Pentagon and government.
4. Interreality and authentication In the Second Life visitors are allowed to populate, build and exploit initially empty spaces. As a result the new reality that is thus created is, remarkably enough not entirely ‘virtual’, but is becoming gradually more linked to our physical reality.
Relationships between social, economical, and psychological status of game players and their respected avatars in the virtual environment are a subject of current research.
5. Motivation As the physical and the virtual worlds come closer to each other, the distinction between the two fade and the need arises for security systems capable of working in the contexts of interreality and augmented reality.
In ‘Architecture of a Cyber Culture’ (2003), Van Kokswijk describes this phenomenon as “the hybrid and absolute experience of physical and virtual reality”.
Unfortunately, currently available biometric systems are not designed to handle visual and behavioral variations observed in non-human agents and consequently perform extremely poorly.
6. Preliminary work Biometric and intelligent robots/software:
1998 M.J. Lyons et.al. published a report: ”Avatar Creation using Automatic Face Recognition”, avatar to be created almost automatically from the human face – biometric synthesis
Advanced Research Projects Activity is to develop systems to observe avatars to study how they behave and communicate with one another for insights into how real-life people
7. What We Did ? Collected avatar, bots and robot image database using manual and automatic approaches
Presented strong case for Artimetric research as a new area of science through over 200 references
Applied concept of biometric feature based recognition to avatar face classification
Postulated that multimodal biometric system architecture can support both visual and behavioral pattern recognition
8. Database
9. In other words...
We saved a bunch of images of virtual faces.
In even simpler terms...
These guys got their pictures taken.
Say Cheese! In other words, we took a bunch of pictures of virtual people's facesIn other words, we took a bunch of pictures of virtual people's faces
10. Why We Did It?Artimetrics: Research aimed at visual and behavioral recognition and identity verification of industrial and personal robots, virtual world avatars and other non-biological agents is Artimetrics.
Extension of biometrics
As the non-biological entities become common a need arises to authenticate them just like people 10
11. Second step: Development of avatar recognition system based on biometric principles
Robot/avatar emotion understanding
Collection of other robot-data such as:
Chat-log
Synthetic voice recordings
Game strategy logs
11
12. Application to avatar recognition: feature based
13. Emotion recognition Emotion recognition method is based on expression change analysis and morphing expression images in 2D and 3D based on the powerful computational geometry methods
It provides a simple fully automated algorithm based on the distance transform that computed the mapping between pixels in a monochrome image through a clever process of sweeping the image and reusing the information obtained on the previous step, utilizing Sibson coordinates and Delaunay triangulation.
17. Third step: 11/22/2011 17
18. 11/22/2011 18
19. 11/22/2011 19
20. 11/22/2011 20
21. 11/22/2011 21
22. 11/22/2011 22
23. 11/22/2011 23
24. Thank you!