Many enterprises connect to the internet through a proxy server as a standard practice. But security researcher Jerry Decime recently discovered a man-in-the-middle (MitM) attack that exploits proxy use.
He gave the exploit a snazzy name, FalseCONNECT, and got a CVE tracking number for it. But what is this software vulnerability all about?
Bad Connection
Decime demonstrated a flaw in how applications from major vendors (excluding Lenovo) respond to HTTP CONNECT requests via HTTP/1.0 407 Proxy Authentication Required responses. The major vendors affected by this flaw include Apple, Microsoft, Oracle and Opera. Other vendors are still in the process of evaluating the situation as it pertains to them.
The vulnerability note summarized the inner workings of the issue rather nicely, if densely. An attacker with the ability to modify proxy traffic, it explained, may phish for credentials by forcing the use of 407 Proxy Authentication Required responses.
“WebKit-based clients,” the note said, “are vulnerable to additional vectors due to the fact that HTML markup and JavaScript are rendered by the client Document Object Model (DOM) in the context of the originally requested HTTPS domain.”
In other words, there is a problem with the CONNECT request that affects everyone, and there are also additional problems, such as arbitrary HTML and JavaScript injection, for WebKit-based clients.
An Impractical Software Vulnerability
The attacker must be able to modify proxy traffic to carry out this exploit. That requires extensive preparation as well as decent execution skills. An unskilled cybercriminal looking for a quick buck is unlikely to use this type of attack.
Decime also noted that Apple fixed the HTTPS middling flaw in iOS and OS X well within the 45-day disclosure window established by CERT. Other vendors may still have to determine if they are affected and how to remediate any problems.
While it’s important to be aware of this vulnerability, but it’s relatively ineffective if the endpoints and infrastructure of the network have been secured. If solutions such as IDS or IPS exists solely on the network’s exit nodes, however, they may miss issues affecting users on local subnets. Nothing is perfect in security.
While this is a relatively shallow flaw in a trusted and widely used solution, it should remind us to continue to look for new ways to address old problems.