Hi everyone,

there are false reports circulating in MITRE CVE and other security  
sites about a vulnerability in Agavi that leads to an exploit allowing  
remote attackers to read arbitrary files on the server's filesystem.

This vulnerability found by the reporters does, of course, not exist  
in Agavi itself, but is a problem in the application code (and has  
been validated as such) of the website where the vulnerability was  
found.

We have found several vulnerable sites ourselves based on Agavi 0.10  
that have apparently all been created by the same agency in France.

Regardless of the version number, Agavi never gives special treatment  
to specific input parameters and uses them to read files from the file  
system or perform other potentially unsafe operations.

Agavi's unique and consistent input validation concept makes sure that  
developers only have access to data they validated, and in addition to  
shipping build templates that enable the secure "strict" validation  
mode by default, Agavi 1.0 natively assumes this most secure setting  
in code if no mode is given through configuration.
This additional measure further extends Agavi's lead as the most  
secure PHP application framework and complements Agavi's praised  
approach of taking rigorous measures to prevent developers from  
accidentially utilizing unvalidated user input not only from request  
parameters, but also uploaded files, HTTP headers and even cookies,  
drastically reducing the possibility of CSRF and XSS attacks.

We have notified MITRE/DHS, NVD/NIST, SecurityFocus/Symantec,  
milw0rm.org and Sebug.net about the error, as well as the original  
author, "t0fx".

So far, SecurityFocus have corrected the information on their site and  
changed the status to "RETIRED", along with a remark on a sub-page  
about the mistake.

Also, we have been in contact with "t0fx" in the meantime, who reacted  
quickly and sent us the following apology:

> Yes you are right, after some investigations, whe saw that the bug  
> was due to bad filtered values on the websites we tested the vectors  
> on.
> But the vulnerability comes to agavi cms+bad installation..
> I REALLY apologie for the fact that we didn't contact you before  
> posting the exploit on milw0rm.com, but a friend of mine, working  
> with me on finding exploits on websites, told me that he contacted  
> you 2 days before... I asked him again this evening and he told me  
> that he forgot to do it.... We usually never post vulnerabilities  
> unless the coder is contacted, so I'M VERY SORRY FOR THAT.
> We did not wanted to attempt at your reputation you can trust me...

He assured me that he would contact the various sites where the  
vulnerability was posted, and notify them that the provided  
information is incorrect and that the vulnerability does not exist.

Here is a list of the vulnerabilities in various security databases:
- http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2008-4920
- http://web.nvd.nist.gov/view/vuln/detail?vulnId=CVE-2008-4920
- http://www.securityfocus.com/bid/32086
- http://www.sebug.net/exploit/5066/
- http://www.milw0rm.com/exploits/6970

If you know of any other security databases or additional resources or  
people who might need this information, please forward this email  
accordingly.

Greetings from Munich,

- David

_______________________________________________
Agavi Dev Mailing List
[email protected]
http://lists.agavi.org/mailman/listinfo/dev

Reply via email to