|
Posted by Walter Roberson on February 6, 2006, 1:27 pm
If you were Registered and logged in, you could reply and use other advanced thread options >Let's assume we have an online arcade game where the winners earn
>prizes (to attract the bad guys), and let's assume all security aspects
>are addressed except the one that follows. Is it possible to prevent an
>attacker from writing a program to analyze the screen and help him win?
>As an example, if the game is a pacman clone an attacker can analyse
>the screen, and model the entire game: the maze, his position, the
>position of the ghosts, the food, etc, and find an algorithm to help
>him move and win. Any idea?
Study your history: look at "rogue-a-matic".
Essentially, if the game contains enough structure and predictability
for a human to be able to determine a strategy for success, then
it is also possible to write a program that plays the game and
emperically discovers the conditional probabilities.
So what you do is impose a cost-per-play and make the game sufficiently
complex that the cost to play enough games to model the probabilities
would exceed the prize money available.
Also, you can use strategies that make automated analysis difficult,
such as exploiting the human perceptual system to transfer information
that would be easily noticed by humans but would be difficult for
a machine to discover. Investigate "machine vision" for further
information about what is easy or difficult to model in computers.
pacman was particularily easy to solve because it was entirely
predictable -- every game was exactly the same as every other game
in which you made the same moves at the same relative times.
| 
XML site map