Take the 2-minute tour ×
Super User is a question and answer site for computer enthusiasts and power users. It's 100% free, no registration required.

I have the following scenario:

A certain process (A) creates files and I want to write a program which reads these files to generate some statistical data.

The problem is simple. I want to be sure that the files created by (A) haven't been corrupted/faked. So the only one who is allowed to change/modify/create those files is (A). If any other process changed some of the files or creates similar files I want to detect this and mark these file as faked/corrupted. I need to be absolutely sure that these files haven't been changed in any way by another process than (A).

The process (A) is only running on windows system. More specifically windows xp or above.

Is this even possible? Is it enough to watch the filesystem or do I have to inject in some way (A) to get the output data before it's stored in a file?

share|improve this question
    
If you're not the administrator, you can't do this properly. A good start would be to create a separate user, set up permissions on the target file/directory so that only this user can access it, run process A as that user. –  DCoder Aug 21 '13 at 7:49
1  
Who is your adversary? Why should anyone meddle with your files? –  Deer Hunter Aug 21 '13 at 9:23
    
Anyone who wants to fake the statistic –  bakunin Aug 21 '13 at 10:08
    
bakunin, why do you allow them anywhere near your servers/workstations? –  Deer Hunter Aug 21 '13 at 10:09
    
Let's say process A is a game. And I want to generate statistical data. The game can be installed anywhere - so I don't have access to these computers. BUT my program wants to generate the statistical data from the log files the game generates on their computers. So the game and my program is running at the same time at the same computer. Of course the client/gamer uses my program because he wants to see the statistic (which may be shared on a website). But some of those gamers could possibly just fake the log files and therefore fake the statistical data. –  bakunin Aug 21 '13 at 10:20

4 Answers 4

up vote 1 down vote accepted

This is a DRM problem and there isn't really a secure answer. You need to use encryption so that only A can sign the data as authentic, but that requires A to know a secret that can be discovered by the user of the computer, thus they can fake the information.

You could try to have server B somewhere sign the data so that the owner of the system A is running on can't do it, but then how does B know that it is actually talking to a legit program A and not something the user cooked up. There isn't any way to determine this either.

The only real way to do it is to have every important action happen as the result of something on the server as well as all of the criteria that need to be correct to ensure a proper version of the data. This is why server side logic is used a lot in MMOs and some other online games to prevent cheating. If you don't own and control the hardware, you can't trust it. DRM tries to, but it is a losing battle unless we get to a point where the user that owns the system is just a "guest" of some third party that really runs the system, and that's a very scary world for many other reasons.

share|improve this answer
    
nice answer - thanks for the detailed explanation! - meh can't upvote it... –  bakunin Aug 21 '13 at 13:45

There is a security model for that it is called Clark-Wilson model. Only single app is allowed to change the data.

What I would do is I would create files with digital signature on them, and some freshness checks. That way your application B can be certain that files were written by A and not manipulated by any other process. Plus checking for freshness helps you mitigate replay attacks.

share|improve this answer
    
That's interesting.. can you provide more details please? Am I right, that you would create a hash for each file? The problem is, the file may be changed by process A at some time (append data). So the hash value would change every time. So I have to ensure, that (as you say) only process A can change/create the files. But I have actually no idea how to do that. When working with the right system, the user may change the rights also. –  bakunin Aug 21 '13 at 12:17
    
So do not sign files only sign data in files, ie. use xml and sign every xml part that contains some atomic data. That way your A will only have to hash and sign appended data. –  fatfredyy Aug 21 '13 at 13:08
    
I have no access to the code of A, it's not my application. –  bakunin Aug 21 '13 at 13:14
    
If not then you have to use operating system level access control to allow only software to modify and read those files. –  fatfredyy Aug 21 '13 at 13:44

I would probably make sure that the application that generates the files also creates an MD5 hash sum of the file. Store that hash sum to a seperate file somewhere.

At a later point, probably when you need to make use of the file, you can rehash the file and ensure the MD5 sum still matches.

share|improve this answer
    
I have no access to the code of process A. Otherwise it would be easy anyway, wouldn't it? –  bakunin Aug 21 '13 at 12:24
    
Ah, I see. In that case my instinct is that it would seem impossible to determine if a file was created by 'process A' or 'process X', and similarly hard to determine if it had been modified by an authorised process. –  PaulG Aug 21 '13 at 12:29
    
I'll give it a try. Impossible sounds quite unsatisfactory... –  bakunin Aug 21 '13 at 12:45

Microsoft offers a free File Checksum Integrity Verifier for such purposes.

Information copied from the readme file:

1.What is fciv?

Fciv is a command line utility that computes and verifies hashes of files.

It computes a MD5 or SHA1 cryptographic hash of the content of the file. If the file is modified, the hash is different.

With fciv, you can compute hashes of all your sensitive files. When you suspect that your system has been compromised, you can run a verification to determine which files have been modified. You can also schedule verifications regularily.

2.Features:

  • Hash algorithm: MD5 , SHA1 or both ( default MD5).
  • Display to screen or store hash and filename in a xml file.
  • Can recursively browse a directory ( ex fciv.exe c:\ -r ).
  • Exception list to specify files or directories that should not be computed.
  • Database listing.
  • hashes and signature verifications.
  • store filename with or without full path.

3.Syntax:

Usage:  fciv.exe [Commands] <Options>

Commands: ( Default -add )

   -add    <file | dir> : Compute hash and send to output (default screen).

      dir options:
      -r       : recursive.
      -type    : ex: -type *.exe.
      -exc file: list of directories that should not be computed.
      -wp      : Without full path name. ( Default store full path)
      -bp      : base path. The base path is removed from the path name of each entry
   -list   : List entries in the database.

    -v     : Verify hashes.
           : Option: -bp basepath.

    -? -h -help : Extended Help.

Options:
        -md5 | -sha1 | -both    : Specify hashtype, default md5.
        -xml db                 : Specify database format and name.

To display the MD5 hash of a file, type fciv.exe filename

Compute hashes:

fciv.exe c:\mydir\myfile.dll
fciv.exe c:\ -r -exc exceptions.txt -sha1 -xml dbsha.xml
fciv.exe c:\mydir -type *.exe
fciv.exe c:\mydir -wp -both -xml db.xml

List hashes stored in database:

fciv.exe -list -sha1 -xml db.xml

Verifications:

fciv.exe -v -sha1 -xml db.xml
fciv.exe -v -bp c:\mydir -sha1 -xml db.xml

4.Database storage format:

xml file.

The hash is stored in base 64.

5.Verification:

You can build a hash database of your sensitive files and verify them regularily or when you suspect that your system has been compromised.

It checks each entry stored in the db and verify that the checksum was not modified.

share|improve this answer
    
I don't see how this could help me. I need to be sure that (A) created/modified the file. Also (A) is appending data to the log files. So I guess, a hash value isn't a solution? –  bakunin Aug 21 '13 at 13:28

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.