Date: Mon, 07 Jun 2010 11:07:54 -0400 From: Larry Adams <thewitness@...ti.net> To: "Steven M. Christey" <coley@...us.mitre.org> CC: Jan Lieskovsky <jlieskov@...hat.com>, oss-security@...ts.openwall.com, Nahuel Grisolia <nahuel@...sai-sec.com>, Stefan Esser <stefan.esser@...tioneins.de>, Cacti Developers <developers@...ti.net>, Tony Roman <roman@...order.com> Subject: Re: CVE Request -- Cacti v0.8.7 -- three security fixes On 6/7/2010 10:21 AM, Steven M. Christey wrote: > > On Tue, 1 Jun 2010, Jan Lieskovsky wrote: > >>> [C], SQL injection and shell escaping issues reported by Bonsai >>> Information Security (http://www.bonsai-sec.com) >>>  >>> http://www.bonsai-sec.com/blog/index.php/using-grep-to-find-0days/ >>>  >>> http://www.bonsai-sec.com/en/research/vulnerabilities/cacti-os-command-injection-0105.php >>> >>> >>> >>> ... >>> > >> 2, OS command injection issue, CVE-2010-1645 / BONSAI-2010-0105 >> References:  >> http://www.bonsai-sec.com/en/research/vulnerabilities/cacti-os-command-injection-0105.php >> >> Proper patches are the following three: (noticed by Tomas Hoger >> && confirmed by Tony Roman, thanks for it!) >>  http://svn.cacti.net/viewvc?view=rev&revision=5778 >>  http://svn.cacti.net/viewvc?view=rev&revision=5782 >>  http://svn.cacti.net/viewvc?view=rev&revision=5784 > > The BONSAI-2010-0105 references two problems, one for ping.php and > another one having to do with a "Vertical Label" in a "Graph Template." > > I don't see evidence of this vector in the revisions listed above. > Does anybody else? > > (If the "Vertical Label" issue went unpatched, then a separate CVE > should probably be assigned to it.) > > - Steve Steve, I just validated that the Vertical Label via the hostname field injection is not resolved from a UI perspective in host.php. I have not checked lib/rrd.php for proper escaping and will do that shortly. Generally, we're not as concerned when it comes to components of the UI that are not accessible for the guest account. However, I will close this loophole today and prior to Cacti 0.8.7g beta. Simply put, we should not permit any name there that is not either an IP address or something that is conforming (aka hostname or fqdn). I toyed with the idea of a gethostbyname validation. However, there are cases in Cacti where we have hosts that are represented more as Objects and not as physical hosts (aka no avail check), so it will have to be a hostname that is syntactically correct. I'll get you a commit # once it's in. Regards, Larry Adams Lead Cacti Developer
Powered by blists - more mailing lists
Please check out the Open Source Software Security Wiki, which is counterpart to this mailing list.