What is the real risk of the Shadow Devices we described back in our first post? It is clear that more organizations don’t really take their risks seriously. They certainly don’t have workarounds in place, or proactively segment their environments to ensure that compromising these devices doesn’t provide opportunity for attackers to gain presence and a foothold in their environments. Let’s dig into three broad device categories to understand what attacks look like.
Do you remember how cool it was when the office printer got a WiFi connection? Suddenly you could put it wherever you wanted, preserving the Feng Shui of your office, instead of having it tethered to the network drop. And when the printer makers started calling their products image servers, not just printers? Yeah, that was when they started becoming more intelligent, and also tempting targets.
But what is the risk of taking over a printer? It turns out that even in our paperless offices of the future, organizations still print out some pretty sensitive stuff, and stuff they don’t want to keep may be scanned for storage/archival. Whether going in or out, sensitive content is hitting imaging servers. Many of them store the documents they print and scan until their memory (or embedded hard drive) is written over. So sensitive documents persist on devices, accessible to anyone with access to the device, either physical or remote.
Even better, many printers are vulnerable to common wireless attacks like the evil twin, where a fake device with a stronger wireless signal impersonates the real printer. So devices connect (and print) documents to the evil twin and not the real printer – the same attack works with routers too, but the risk is much broader. Nice. But that’s not all! The devices typically use some kind of stripped-down UNIX variant at the core, and many organizations don’t change the default passwords on their image servers, enabling attackers to trigger remote firmware updates and install compromised versions of the printer OS. Another attack vector is that these imaging devices now connect to cloud-based services to email documents, so they have all the plumbing to act as a spam relay.
Most printers use similar open source technologies to provide connectivity, so generic attacks generally work against a variety of manufacturers’ devices. These peripherals can be used to steal content, attack other devices, and provide a foothold inside your network perimeter. That makes these both direct and indirect targets.
These attacks aren’t just theoretical. We have seen printers hijacked to spread inflammatory propaganda on college campuses, and Chris Vickery showed proof of concept code to access a printer’s hard drive remotely.
Our last question is what kind of security controls run on imaging servers. The answer is: not much. To be fair, vendors have started looking at this more seriously, and were reasonably responsive in patching the attacks mentioned above. But that said, these products do not get the same scrutiny as other PC devices, or even some other connected devices we will discuss below. Imaging servers see relatively minimal security assessment before coming to market.
We aren’t just picking on printers here. Pretty much every intelligent peripheral is similarly vulnerable, because they all have operating systems and network stacks which can be attacked. It’s just that offices tend to have dozens of printers, which are frequently overlooked during risk assessment.
If printers and other peripherals seem like low-value targets, let’s discuss something a bit higher-value: medical devices. In our era of increasingly connected medical devices – including monitors, pumps, pacemakers, and pretty much everything else – there hasn’t been much focus on product security, except in the few cases where external pressure is applied by regulators. These devices either have IP network stacks or can be configured via Bluetooth – neither of which is particularly well protected.
The most disturbing attacks threaten patient health. There are all too many examples of security researchers compromising infusion and insulin pumps, jackpotting drug dispensaries, and even the legendary Barnaby Jack messing with a pacemaker. We know one large medical facility that took it upon itself to hack all its devices in use, and deliver a list of issues to the manufacturers. But there has been no public disclosure of results, or whether device manufacturers have made changes to make their devices safe.
Despite the very real risk of medical devices being targeted to attack patient health, we believe most of the current risk involves information. User data is much easier for attackers to monetize; medical profiles have a much longer shelf-life and much higher value than typical financial information. So ensuring that Protected Health Information is adequately protected remains a key concern in healthcare.
That means making sure there aren’t any leakages in these devices, which is not easy without a full penetration test. On the positive front, many of these devices have purpose-built operating systems, so they cannot really be used as pivot points for lateral movement within the network. Yet few have any embedded security controls to ensure data does not leak. Further complicating matters, some devices still use deprecated operating systems such as Windows XP and even Windows 2000 (yes, seriously), and outdated compliance mandates often mean they cannot be patched without recertification. So administrators often don’t update the devices, and hope for the best. We can all agree that hope isn’t a sufficient strategy.
With lives at stake, medical device makers are starting to talk about more proactive security testing. Similarly to the way a major SaaS breach could prove an existential threat to the SaaS market, medical device makers should understand what is at risk, especially in terms of liability, but that doesn’t mean they understand how to solve the problem. So the burden lands on customers to manage their medical device inventories, and ensure they are not misused to steal data or harm patients.
Industrial Control Systems
The last category of shadow devices we will consider is control systems. These devices range from SCADA systems running power grids, to warehousing systems ensuring the right merchandise is picked and shipped, to manufacturing systems running robotics, and heavy building machinery. All these devices are networked (whether directly or indirectly) in today’s advanced factories, so there is attack surface to monitor and protect.
We know these systems can be attacked. Stuxnet was a very advanced attack on nuclear centrifuges. Once within the nuclear facilities network, the adversaries compromised a number of different types of control systems to access centrifuges and break them. In a recent attack on a German blast furnace, the control systems were compromised and general failsafes were inoperable; the facility went offline while they cleaned the systems up.
In both cases, and likely many others that aren’t publicized, the adversaries are very advanced. They need to be – to attack a centrifuge like Stuxnet you need your own centrifuges to test on, and they aren’t exactly easy to find on eBay. You cannot just load a blast furnace into the pick-up Saturday morning for a pen-test.
That may comfort some people, but it shouldn’t. The implication is that control system defenders aren’t dealing with the Metasploit crowd, but instead trying to repel well-funded and capable adversaries. They need a very clear idea of what their attack surface looks like, and some way of monitoring their devices; they cannot rely on compliance mandates to require advanced security on their control systems.
Another consideration with control systems is the brittle nature of many of them. They are hard to test because you could bring down the system while trying to figure out whether it’s vulnerable. Most organizations don’t like that trade-off, so they don’t test directly. This means you need indirect techniques – definitely to figure out how vulnerable they are, and probably to discover and monitor them as well.
Which makes a good segue to our next post, where we will dig into two aspects of protecting shadow devices: Visibility and Control. First and foremost you need to figure out where these devices really are. Then you can worry about how to ensure they are not being attacked or misused.