| home | help | back | first | fref | pref | prev | next | nref | lref | last | post |
To: www-security@ns1.rutgers.edu Date: Fri, 19 Aug 1994 00:09:15 EDT From: Marc Horowitz <marc@MIT.EDU> >> System security things should be statically linked. Especially on >> Suns, with their LD_PRELOAD functionality. The load library path >> remains a vulnerability. No it doesn't. If the program runs as the user, then if I can spoof your libraries, I can just as easily spoof ls as a "system security thing". If the program is setuid, then the dynamic loader should ignore LD_LIBRARY_PATH if euid != ruid. This is what SunOS (at least the newer versions) does. >> Of course you could counter with 'I am only going to allow the >> authentication methods listed in this file', but then there is no >> advantage in dynamic linking over runtime linking, and very little >> over static linking. This is *exactly* what I would say. The advantages over static linking are clear: you don't have to do it when you get a new security mechanism. Dynamic linking also requires you recompile if you want to add new mechanisms, as the dynamic linking systems I'm familiar with assume one symbol space; if you have two libraries with the same function names, you can't use both at the same time. Now that I've said that, I'll be a hypocrite. This is www-security, not random-os-security. There are other lists for that. Marc
| home | help | back | first | fref | pref | prev | next | nref | lref | last | post |