Our SAGs: Design Aspects and Improvements - Semantic Acquisition Games: Harnessing Manpower for Creating Semantics - page 127

8.2.4 Cheating Vulnerability
Just as the Little Search Game, the PexAce also suffered from several cheating
vulnerabilities (though it must be said that this was the case of the game publicly
deployed over general images). One way to bypass the proper playing, was to use
an automated approach—a wrapper that was playing the game instead of a human
player. The robot would simply sequentially flip all of the cards, “remember” their
positions and bitmaps and find pairs by comparing the bitmaps. One of the game
players managed to program such robot. It led to games being solved in virtually
no time (naturally, without entering a single annotation), leading to almost perfect
scores.
To reduce the use of robots we implemented the following:
1. We displayed slightly different images as pairs. This was achieved through slight
blurring, pixel color mixing, toning, application of watermarks and slight card
rotation. To a human eye, the images looked almost the same (and at the same
time, the modifications were not disturbing) but for the simple pixel-by-pixel
comparison, the task was suddenly impossible. Of course, we knew that some
players might event come with more sophisticated methods of image comparison.
However, no such attempts were made since.
2. We also applied the a posteriori cheating detection heuristics, which we used
with the LSG. As indicators of dishonest gaming, we seek for co-occurrence of
the following: (1) short gameplay times, (2) high ladder ranks, (3) a gameplay
with no errors (repeated card flips) and (4) no created annotations for any of the
images.
In case of PexAce, the combination of the reasonable prevention along with the
application of a posteriori cheating detection (done permanently by a daemon script
running on the game server) successfully distracted the dishonest players from abus-
ing the game.
8.3 CityLights
The CityLights too is a game designed to be aware of the cold-start problems—it is
a single player game. Its purpose is unique. Unlike other SAGs, this one do not create
new metadata, but validate the existing ones. It is fueled by the existing (musical)
resource plus metadata dataset , which the players are correcting and at the same
time, getting score based on. Its design involves several notable features and also
allows certain generalizations.
Several of our design aspects played a minor role in the design of the CityLights.
The purposeful task of the game is to validate individual tag assignments to music
resources. This is a naturally small task and we had no need in specially decomposing
it. The tasks can also be considered equally difficult , resp. its difficulty depends more