[Systrace] Question regarding the security of systrace

[Johannes Nicolai]

Hi,

I hope, that I do not annoy anybody on this list when I ask what you think 
about:
http://cert.uni-stuttgart.de/archive/bugtraq/2004/03/msg00282.html

I know that Marius has already responded to this claim in:
http://www.monkey.org/openbsd/archive/misc/0403/msg01611.html

I am not a security expert, so I ask you how serious do you think the 
mentioned "bugs" are and if there are any new information about it.

To precise my question: I like to use systrace for the next world championship 
of RealTimeBattle (http://realtimebattle.sf.net) in order to protect the 
machine for potential malicious programs (some competitors only send binary 
code) that do not only act as robots in the game but also try to do harm or 
influence other processes.

The robots can only execute a few system calls, communicate with their team 
mates via Unix Domain Sockets in a special directory and access configuration 
files in this directory.

They won't be allowed to fork or to ptrace (exevcve is okay).

I wonder whether one can exploit systrace if only this limited set of system 
calls is allowed at all.

Perhaps you can help me with this question.

Thanks in advance

Johannes Nicolai