Slashdot Mirror
Malware Is 'Rampant' On Medical Devices In Hospitals
Dupple sends this quote from MIT's Technology Review:
"Computerized hospital equipment is increasingly vulnerable to malware infections, according to participants in a recent government panel. These infections can clog patient-monitoring equipment and other software systems, at times rendering the devices temporarily inoperable. While no injuries have been reported, the malware problem at hospitals is clearly rising nationwide, says Kevin Fu, a leading expert on medical-device security and a computer scientist at the University of Michigan and the University of Massachusetts, Amherst, who took part in the panel discussion. [He said], 'Conventional malware is rampant in hospitals because of medical devices using unpatched operating systems. There's little recourse for hospitals when a manufacturer refuses to allow OS updates or security patches.' ... Despite FDA guidance issued in 2009 to hospitals and manufacturers—encouraging them to work together and stressing that eliminating security risks does not always require regulatory review—many manufacturers interpret the fine print in other ways and don't offer updates, Fu says. And such reporting is not required unless a patient is harmed."
I don't know about medical devices, but I do know that the last time I was in the emergency room I brought my laptop since I knew I would be there for a few hours. After getting tired of games and slashdot I decided to poke around the wifi network that I was on. I found an unsecured smb share on the network and downloaded a 17gb .bak file of patient records. Needless to say I deleted the file and sent an anonymous email to the administrator. 3 months later nothing had changed....
Windows is not intended to be used in life-critical situations such as medical hardware or nuclear reactor control. It's right there in capital letters in the EULA.
Someone's being a cheapskate here and decided to use windows instead of paying to develop a custom medical OS.
Dad has owned an ultrasound service business since the late 70s. My brothers and I all worked for him in varying capacities, before becoming engineers ourselves.
In my experience: the amount of willful ignorance towards all manner of IT in the medical field is nothing short of astounding.
I hate to say it, because I love alot of these people- but I chalk it up to the arrogance of the doctors and administrators. They treat anything IT related on the same level as an issue regarding say, HVAC or sanitation. That is to say, beneath them.
Which is fine, except in this case the "HVAC" can be programmed by a remote intruder to emit Zyklon B.
Everyone would just start leaving hospital with an enlarged wanger, and a $12,000,000,000,000,000 bank deposit from a Nigerian prince.
Ok, I'm only a student. So I don't know anything. But I sorta THOUGHT that the standard for a mission critical system (aka something like a heart monitor, blood gas analyzer, etc etc etc) would be to NOT use any software in your system that you don't have 100% control over.
You know, rather than picking some version of windows, use an embedded linux. Add the bare minimum graphics libraries you need in order to draw a gui. Isolate the threads that actually do the mission critical stuff (say, reading the sensor and displaying the output) from the ones that do other tasks (like handling all the complex menus and the network connectivity and so on). Heck, use a separate physical CPU for the mission critical stuff, and give it it's own dedicated display so that no matter what, it keeps displaying the important data. The hardware to do this is cheap.
And firewalls should be integrated into the devices themselves - even Linux can theoretically catch a worm, and so it should apply strict filtering rules on any communications with the network.
I can fully understand the reluctance of the manufacturers to issue software patches. Building the system so that it's practical to not ever patch it (well, maybe patch it a couple times to eliminate any bugs found after release) is a good thing. Everyone here must know that the best way to break a working machine is to shut it down and change something.
I worked as an IT Manager in a hospital for a few years, and know a little bit about this... The first issue is that these systems typically CAN NOT be upgraded, and this is not due to the MFG not wanting to upgrade, this is a FDA compliance issue... If they upgrade the software, they have to do some very expensive certifications with the FDA, these same certifications delay the release of medical equipment to the point that much of the technology is already close to being outdated when it hits the market.
... but we managed just fine.
Our solution, which seems simple enough, was that every type of medical equipment was located on a different physical network (for critical pt. monitoring equipment) or at a minimum a seperate VLAN on the main network. All network access to this equipment was blocked except for very specific exceptions that were allowed based on the absolute need of that piece of equipment. We had no issues with any of these infections or malware, although it did increase the man-hours overhead especially when working with the vendors that would sometimes wonder why they could not hit the internet from the X-Ray machine
+++ATH0 NO CARRIER
The term medical device has a broad definition; it includes obvious things such as laboratory analysers, X-ray equipment, etc., but it also includes PCs running specific types of software, such as medical records software. Most of these things run general purpose OSs - some embedded; some desktop.
E.g. Windows XP is a common platform for things like ultrasound scanners, MRI scanners, etc. XP embedded is quite common on things like laboratory equipment. Variants of linux are also in widespread use - albeit, often old. E.g. I work with an MRI scanner that runs a 2.2 kernel.
Now, things like analysers and scanners are usually on their own VLAN (or should be) with connections only to their application servers, with the servers heavily firewalled from the general purpose VLANs; however, this often isn't the case, and I've seen a number of installations where you can just sit down at a random PC, and SSH into an MRI scanner (these things usually have generic root passwords which are written in the service manual - once you know what the passwords are, you can get into any device of that make and model).
The biggest problem, however, is that these machines never get updated. The manufacturers often won't support any updates to the OS, or even permit hotfix installation, nevermind a 3rd party security package (for more general purpose devices). For example, one hospital earlier this year, upgraded their PACS system (software for storing and displaying X-ray/MRI/CT images) and bought a new set of dedicated workstations (quad core, Xeon E5, 8GB RAM, Dual Quadro), but because the PACS client software had to interface with a number of other client software packages, and those vendors had strict requirements; these machines ended up being loaded with XP SP1 32-bit and Java 1.4. Unsurprisingly, these aren't regularly patched, and more importantly, they can no longer update their anti-virus software as the current version of their chosen AV software won't run on this configuration (so they're stuck using an obsolete, unsupported version).
I saw an extreme example of this a few years ago when the Confiker worm hit. There were a group of hospitals in a major city, which shared the same infrastructure, and they had a very large PACS system. The worm got onto the PACS VLAN, and essentially killed the servers. The system was completely down for days, because as soon as the servers we rebooted or re-imaged; the worm killed them again. The vendor stubbornly refused to apply the hotfix and refused permission to install the hospital's antivirus system on the servers/workstations. The only thing that got it moving was when the CEO of the hospitals made a conference call with the hospitals lawyers and the CEO of the PACS vendor, telling them that they were going to f**k them so hard with the SLA stick, that they wouldn't be able to sit down for a month. After that call, the vendor agreed to install the hotfix, and the system came back online.
I found an unsecured smb share on the network and downloaded a 17gb .bak file of patient records. Needless to say I deleted the file and sent an anonymous email to the administrator. 3 months later nothing had changed....
Usually anyone who dares tell the Emperor that he's actually naked and not wearing any "new clothes" gets his head chopped off for pointing out the truth.
Lemme tell you what would've happened at one particular hospital I know of: The IT administrator would've contacted law enforcement and provided them with all the video footage from the multitudes of security cameras around the place, along with the patient and visitor lists, as well as all the the wifi access and activity logs containing your mac address and anything else logged and/or identifiable about your laptop, to try to find out your real identity for criminal prosecution purposes.
Despite the fact that they are extremely weak in securing their network resources in the first place nor do they have any realtime alerting mechanisms to detect any kind of unauthorized access while in progress.... they do go to ridiculous lengths to log and record everything necessary to try to identify you so they can come and get you long after the fact.
The question is why would medical devices get malware on them just because the OS is unpatched? The frigging device could be Win95 but it shouldn't matter if all it ever runs is the vendor's software.
If people are browsing the internet on them or sticking USB drives in them they are doing things very wrong.
Medical people should be familiar with the terms "quarantine" and "isolation".
While this should be true, these devices are increasingly being connected to networks to offer integration with EHR/HIS for polling information, and especially in radiology, where images are being sent digitally to PACS. These machines often stay unpatched, yet get connected to the network for transfers. It's important to maintain a separate "medical device" network, but this only goes so far, especially when vulnerabilities bypass the Windows firewall on the medical device, allowing some infected PC/device/server to broadcast worms all over the place.
You're right about the network architecture, but things rapidly get complex.
Let's take the example of MRI/CT. How much data is in a CT or MRI study, or even an X-ray study? A single X-ray image (e.g. a Chest X-ray) taken with a modern digital machine, is about 60MB (30 megapixel image, 16 bits per pixel).
My new CT scanner, if I prescribe a "full neuro" protocol, generates 16000 files of 500 kB each. The reason I'm doing a "full neuro" it means that minutes count. I need to have that data set sent to not just a PACS (image repository and viewing software), but also to a PC with 3rd party software (which has the complex software capable of analysing the data) and I have to have it ready within 5 minutes. Not only do I need to have it in my office in 5 minutes, the doctor who is dealing with the patient in the ER, needs to have (some) of it in the ER within 5 minutes. Then, after everything is said and done, I need to send the data to my office at the university, so that I can run it through my research software.
If it was just PACS - no problem. You put the scanners and the PACS incoming-data server on a restricted VLAN. Have the incoming PACS server communicate with the main PACS application and data-store servers over a private VLAN, and have the PACS app servers face the hospital clients on the main hospital VLAN (or individual departmental VLANs).
However, at my hospital we also get several hundred CTs/MRIs sent in from outside per day, that need to get onto the PACS. Many come on CD/DVD. Some come via VPN tunnels. Some come via 3rd party proprietary transfer services. (The DICOM protocol used to transfer medical images doesn't support encryption, so must be tunnelled in some way). Now you have to somehow connect all these incoming points to your restricted VLAN (or you open your wallet to your PACS vendor for another software license at a cost that makes oracle enterprise look like chump change).
What if your PACS vendor has you buy the balls on your SAN contract, so that you are paying $10 per GB + $2 per GB per year? Do you really want to send that 8GB dataset to PACS (which can't actually do anything useful with it- and remember, as a medical-grade archiving device, you can't delete)? Or do you now need to start putting PCs with 3rd party software on your restricted VLAN so they can talk to the scanners?