Embedded Systems Still Unprotected

Monday, March 5, 2012 @ 02:03 PM gHale

It has been known embedded web servers are an easy mark when it comes to being able to hack into them.

That knowledge has existed for quite a few years. With that knowledge it may be easy to assume companies would move to protect their systems. Wrong.

Patched Hole Doesn’t Stop Attackers
Malware Shifts from Safe to Malicious
Malware Strains Meld by Accident
Rail Hack: Govt. Works with Industry

Embedded web servers (EWS) are just as easy to access now than they were years ago. With multi-function printers or video conferencing systems, there can be serious data leaks: Printers store scanned, faxed and printed files on hard disks and then disclose these often sensitive documents. Video conferencing hardware allows outsiders to monitor rooms remotely or listen to meetings that are in progress, said Zscaler’s Michael Sutton at the RSA Conference in San Francisco.

Sutton wanted to scan a million web servers and create a catalogue of all the embedded web servers he found. His first tests involved Nmap and the Google Hacking Database (GHDB). However, neither tool proved very successful, as Nmap doesn’t detect enough EWS fingerprints and will, therefore, produce useless device information. Google, on the other hand, doesn’t allow search queries via scripts and would have required time-consuming manual scans.

The security researcher ended up using the Shodan online scanner. Sutton said Shodan has a huge database containing the HTTP header information of EWS systems, allowing such devices to undergo identification with accuracy. The researcher entered typical character strings from the embedded web servers’ web pages into Shodan. To automate the process, Sutton used a Perl script that only sent HEAD queries via Shodan. The script hosted on several EC2 micro instances in Amazon’s cloud which, according to the researcher, only cost a few dollars.

The scan managed to examine the targeted one million web servers in a short time and came up with the following results: Thousands of multi-function devices (more than 3,000 devices by Canon, 1,200 Xerox photocopiers, 20,000 Ricoh devices, among others), 8,000 Cisco IOS devices and almost 10,000 VoIP systems and phones didn’t require any log-in authentication. The latter included 1,100 devices by the German manufacturer Snom. These devices include packet tapping features and PCAP tracing by default. Imported into Wireshark, the trace can convert into a sound file of the telephone conversation.

The majority of the detected devices did not enjoy password protection, Sutton said. This means that any web user can access their web interfaces through a browser and view the documents stored on such photocopiers and printers, forward incoming faxes to an external number, or record scan jobs. With HP devices, a script can carry out such intrusions that every second calls a URL whose only variable is UNIX epoch time, which is easy to figure out.

The scan run by Sutton also identified more than 9,000 video conferencing systems by Polycom and Tandberg (now Cisco). The most likely reason why these devices were openly accessible on the net is they all use the H.323 protocol and require numerous open ports in the firewall. Sutton thinks administrators shy away from this, placing their systems in a DMZ instead. The IT security expert used a video to demonstrate how he managed to monitor the targeted conference rooms via an accessible video conferencing system that provided sound and images.

Sutton’s company is now providing the brEWS scanner free of charge, which specializes in detecting embedded web servers. To avoid placing the weapon into the hands of criminals, scans can only be run in a /24 subnet. At a later stage, the researcher also plans to offer a browser add-on that will allow administrators to examine protected internal networks; this add-on will carry out the scan and then send the results to the brEWS server for identification.

Leave a Reply

You must be logged in to post a comment.