I am looking at a 2008 academic test paper from a university on secure computing. I could not think of any solution to on particular question.

*******************************************************************
Basically a user has a file in a directory and there is another directory nested inside which is owned by root. The nested directory has a file which is owned by root and user cant access it.

There is a root program which user can invoke which basically checking the permission on user file and creates a symbolic link to it for a few moments and deletes it after use.

This program has a flaw and user has found a way to access the secret file. what could be wrong?
*******************************************************************

Clearly the root program has a race condition vulnerability but how can a user actually exploit it to gain access to root file? This is what is puzzling me. I am attaching the screen shot of the complete question.

Any suggestions would help.

Attached Thumbnails

 

If it is a symbolic link, versus a hard link, I don't see how the root-owned file would be made accessible ...

Even a hard link could not work. The access permissions are part of the file's inode, and are independent of how that inode is reached. A hard link could work around directory permissions, but not file permissions. A symbolic link can do neither.

Gotta luv the web! http://www.cis.syr.edu/~wedu/Teachin..._Condition.pdf

Yes I have gone through this. It exploits the race condition and symbolic links to write to an inaccessible file because the root program which is running is writing to the file and it can be tricked to redirect its input to some other file but in this case the root program is merely creating a symbolic link not opening the file.

The most an attacker could do is to create a symbolic link of file in folder C to imitate his image file apart from that nothing. So how can he gian access to the secret file?

As directory B has 755 permissions, the attacker can exploit the race condition by creating a symlink named "funny.jpeg" to the "secret_records" file.

So the program checks if "the user can view the file" by writing to the file? That doesn't seem like a plausible error. I mean, even in normal non-attacker scenarios wouldn't that cause obvious problems? Maybe that is the intended answer, but IMO it doesn't make much sense.

An alternative presentation with more ideas on potential exploitation techniques. https://www.cs.utexas.edu/~shmat/cou...ou-porter.pptx

A symbolic link is "a tiny file containing a file name." It is resolved as a filename – effectively, the symlink is shorthand.