IEEE Robotics & Automation Magazine - December 2012 - 49

risks, including international terrorist groups and local groups
of radicals. Virtual worlds can be used to create an exact replica of a real-world target and can be utilized to rehearse an
entire attack online, including monitoring the response and
ramifications [78]. We can further illustrate the problem by
analyzing the examples of news reports about the crimes
reported to be committed in the virtual communities. These
crimes are either committed by members of the virtual communities through their avatars or directly by creating malicious software that can accomplish committing the crime. In
either case, "as in the real world, one of the central difficulties
is establishing the identity of individuals" [79]. The examples
given below are by no means exhaustive, because almost any
type of real crime has a virtual equivalent [123]
1) Theft of Virtual Property-"The Netherlands teen sentenced for stealing virtual goods" [33]
2) Virtual Prostitution, Strip Clubs, and Pornography-
"Escorts, the Second Life equivalent of phone-sex operators or prostitutes, are quite common in Second Life" [115]
3) Virtual Gambling-"FBI checks gambling in Second Life
virtual world" [84]
4) Virtual Money Laundering-"Second Life and other online
sites targeted by criminals" [105]
5) Virtual Fraud-"...the 'bank' vanished, and depositors say
their money did, too" [106]
6) Identity Theft-"Second Life charges for real names,
increases identity theft risk" [113]
7) Illegal Content (Child Porn)-"Second Life 'child abuse'
claim" [9].
In addition to numerous examples of virtual crime, it is
also interesting to look at other scenarios in which it would be
useful to track an individual between the real and virtual
worlds. For example, a number of cases have been reported in
which a wanted criminal is easily found in the virtual world
and even taunts authorities by posting status updates and
pictures of his real environment [97], [111].
In the context of investigating criminal and terrorist activity
outlined above, we see six (four nonsymmetrical) scenarios
requiring an automated matching algorithm (see Figure 1). For
each scenario, we have provided a realistic example meant to
motivate the need for a particular matching algorithm [123].
1) Matching a human face to an avatar face and vice versa
[Figure 1(a)]: This capability is useful to connect a person's
real identity to their virtual persona. It is increasingly common to upload a real photograph to serve as a prototype
for a 3-D avatar, as well as to create drawings closely resembling the actual person to serve as the online persona.
Example scenario 1 (avatar to human): During a forensic investigation of a personal computer, a number of
images depicting virtual pornography are found. It is
desirable to run the virtual faces against the database or
real-life sex offenders to see whether any quality matches
can be detected for further investigation.
Example scenario 2 (human to avatar): To follow up a
convicted sex offender forbidden from interactions with
anyone under 18, it might be valuable to do a visual search

of virtual playgrounds to see whether the person is violating the court order in cyberspace.
2) Matching the face of one avatar to another avatar
[Figure 1(b)]: This capability is useful for continuously
tracking a virtual persona through cyberspace at different
times and in different places.
Example scenario 1: An intelligence agency might be
interested in automatically tracking a suspected terrorist
across the virtual community for many days to establish his
contacts and frequently visited places.
Example scenario 2: The same capability might be
extremely useful in personalization and customization of
services, for example, to load certain user preferences if a
recognized avatar returns to a previously visited business.
Example scenario 3: The ability to recognize avatars and
profile them on the basis of appearance is also very useful in
marketing; for example, a cosmetics company selling their
product in a virtual world might be interested in advertising
to all female avatars who appear to use a lot of lipstick.
3) Matching an avatar's face from one virtual world to the same
avatar represented in a different virtual world(s) [Figure
1(c)]: A recent development in the world of virtual communities is the desire to interconnect different virtual
worlds. "One such world, called HiPiHi, is being created in
China. HiPiHi founders said they want to create ways for
avatars to travel freely between its virtual world, Second
Life and other systems-a development that intelligence
officials say make it doubly hard to track down the identity
of avatars" [79].
Example scenario: A well-known criminal has set up
recruitment camps in multiple virtual worlds (Second
Life, Entropia Universe, etc.). To fully understand his
minute-by-minute activity, network of contacts, and
overall strategy, it is necessary to track him beyond a
single virtual world to all corners of the cyberspace.
Because the same real-world photograph will be translated to slightly different-looking avatars depending on
the algorithm used by a specific virtual world, it is very

(a)

(b)

(c)

(d)

Figure 1. Four possible avatar-matching scenarios useful in
forensic investigations [123]. Images from (a) cyberextruder.com,
(b) SecondLife.com, (c) SecondLife.com and EntropiaUniverse.
com, and (d) RedFieldPlugins.com.

december 2012

*

Ieee rObOTIcS & AUTOmATION mAGAZINe

*

49
http://www.cyberextruder.com http://www.SecondLife.com http://www.SecondLife.com http://www.RedFieldPlugins.com
Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - December 2012
