Ask Slashdot: How Do You Handle Hardware That Never Gets Software Updates? (hpe.com)
New submitter pgralla writes from a report via HPE: Many devices, designed for both long-term and short-term use, were shortsighted when it came to flexibility. How do you handle the hardware that never gets software updates, such as embedded systems and task-dedicated equipment? The article that pgralla shared provides the example of medical devices running Windows 7. "Many of the current generation, when they were first released, used Windows 7, and the devices still work well enough that they remain in service today," reports HPE. "But Microsoft ended mainstream support for Windows 7 back in January 2015, so the operating system gets updated only with an occasional security patch as part of Microsoft's extended support. In January 2020, that extended support will end as well." Many IoT devices are in a similar boat as they're powered by embedded Linux and are not designed to be updated after they enter service."
Of course, these outdated devices create all sorts of security concerns. "Hackers and their access to knowledge and computing power only go up as the years pass, which means that long-lived, fixed-firmware devices become ever more insecure over time," says Michael Barr, founder of the Barr Group, which provides engineering and consulting services for the embedded systems industry. The WannaCry ransomware hack in 2017 affected not just PCs but also medical devices, and ended up costing businesses $4 billion.
Of course, these outdated devices create all sorts of security concerns. "Hackers and their access to knowledge and computing power only go up as the years pass, which means that long-lived, fixed-firmware devices become ever more insecure over time," says Michael Barr, founder of the Barr Group, which provides engineering and consulting services for the embedded systems industry. The WannaCry ransomware hack in 2017 affected not just PCs but also medical devices, and ended up costing businesses $4 billion.
....don't buy it.
I've seen SO many people whining about MS' forced reboots, etc. STOP!
If there is not a sensible option available, demand that your vendor make a version that can be sensibly updated. Too many purchasing decisions just don't have any sensible criteria. ("Oh, it's built on Win XP and you aren't updating it? OK - scratch!")
Many old tools are computer based
Some old CNC machines run on MS-DOS and a 286 processor
As long as the hardware stays alive, they continue to do the job
If they must be networked, restrict their access to the local net
I have a number of Rohde and Schwarz FSEB and FSEA spectrum analyzers. These cost at least $80,000 new (I bought them used for a few thousand at most). They come with an old version of windows. I similarly have other electronic test equipment with old Windows or even old Linux which the manufacturer doesn't update any longer. For the Linux-based ones I could hack in a new Linux and make it use the old ABI, forget about Windows.
But what really clued me in was that the Rohde and Schwarz equipment had a battery soldered on the CPU board, and it was an hour-and-a-half service to get to it. A lot of stuff had to be removed.
Similarly, my Tektronix 500-series oscilloscopes had two 40-pin DIP Dallas Semiconductor battery-backed memory and clock chips. The batteries in these die and they aren't socketed. When the batteries die, the 'scopes lose their calibration. The company won't give you the program to recalibrate them.
The manufacturers just want you to buy new ones.
So, obviously I back SDR-based test equipment that's Open Source. Who needs a company that wants to screw you?
Bruce Perens.
Most dedicated systems like this does not belong on the internet, period. So unless there is some flaw or feature need, don't update and it will still work exactly as it did yesterday. And the day before, and the day before that.
Mechanical systems that keep, for example, trains from running into one another by tripping their brakes into full on, are well-understood. I took a course on doing the same thing in mixed hardware-software systems, so it's eminently possible.
The gotcha is you have to keep it really simple and run a validator like spin on it's protocol.
Most developers can do the spin part, but KISS? Distinctly less likely (;-))
davecb@spamcop.net
Implement a firewall with a small microcontroller with a relatively secure TCP/IP stack (ejip if you don't want to spend money, HCC embedded if you do) and do protocol level sanity checking and filtering of all network inputs.
How Do You Handle Hardware That Never Gets Software Updates?
Very carefully. (Buh-DUM-Tshhhh)
Borrowed from “How do porcupines make love?”
With apologies.
Our reign has gone on long enough. Indeed. Summon the meteors.
A basic principle of security is least privilege. If a piece of outdated equipment needs to send udp packets on port 411 to a monitoring station, you set the firewall to allow it to send udp on port 411 to that particular station, and nothing else. If it doesn't need to take to web servers, you don't let it talk to web servers. You allow it to do only exactly what it needs to do.
Not sure what your equipment needs to do? You could check the manual, and otherwise open up Wireshark and set the filter to the IP of the equipment. Have a look at what it is sending and receiving. Then set the firewall to allow only exactly what is needed.
This is also an area where vlans come in very handy. Vlans act like completely separate networks, but they are configured within your switch, so a single 48-port switch can handle a dozen different, totally separate vlans.
Perhaps different parts of your network should be mostly separate, but you need to allow a little bit of specific communication between two vlans. That's when you plug a router or firewall into both vlans and set it to route only specifically allowed traffic between them. This doesn't even require two network ports - the same port can be in multiple vlans and the router can control traffic between vlans issuing a single cat6 cable. This is called "router on a stick".
If some of this went over your head, here's the simple version'
Call someone who has a CCNA Security certification or better (CCNP Security or CCIE Security). Tell them you're thinking about segregating different vlans and using an internal firewall to strictly control internal traffic. They'll get you set up.
Open source has taken off like wildfire. You are using it right now. You are just too dumb to realize it.
I use Slackware, along BSD, financially support projects that I use, and have followed the Linux community since Linus was still in college. It always amazes me how clueless the FOSS community is regarding issues such as this.
Just use Linux...
That's your fault for using M$..
etc.
For regulated systems, especially in pharma manufacturing, you are told what to use, how to use it, when to upgrade it, how to upgrade it, etc. Basically, once the system is certified by the FDA - you don't touch it - PERIOD. You purchase enough compute/control systems when you install it to last you through your production, which could be - 10, 15, 20+ years.
There is no, well, just upgrade to x - it's not allowed.
Before some equally clueless libertarian pinhead starts spouting off about 'over regulation' - stop and think for just one second what this system does. It controls the valves, temperatures, mixing, fermenting, refining, etc. of a chemical that people are to ingest. Where the difference between good and bad is measured in ppm, ppb, or even ppt depending on what's being made. Some endocrine chemicals are measured in 1/10ths or 1/100th of a ug!
Do you really want to apply patches to a system such as this? Doesn't matter that they are 'network', or 'mouse driver', or 'display' - the risk is WAY TOO GREAT to jack around with them.
Keep in mind that 'upgrades' require a new certification of that system, or depending on what it does, the entire production chain - which could run you a couple 10's of millions dollars.
So, before starting the typical FOSS rant, please have a clue of what you are talking about, first.
>ANY device can be infected with a new exploit whether it's up to date or not. New fully updated equipment is no less of a risk than old out of date equipment.
Those are two very different statements. Yes, any device can be compromised by a new exploit - that's kind of the point of developing NEW exploits. But an outdated device can be compromised by a massively long list of well-known exploits - making it far more vulnerable. New exploits are generally financially valuable assets horded by those who know of them, and they will usually be rendered useless shortly after they become public knowledge. Fewer people attacking, means lower risks that you'll be attacked.
I don't know if it's still the case, perhaps the target is no longer as appealing, but I recall that back towards the end of Windows XP's product life, even before 7 came out, the rule of thumb was that a freshly installed copy of (non-updated) XP would be compromised within 20 minutes of being connected to the internet - considerably faster than most people could download the updates necessary to secure it. Not that it was ever 100% secure, but there's a huge difference between going into battle in imperfect armor, and going in wearing nothing but a giant bullseye painted on your chest.
--- Most topics have many sides worth arguing, allow me to take one opposite you.
That doesn't help when a particular device from a particular manufacturer contains non-free software, as do the substitute devices from all competing manufacturers.
Devices running Windows XP are already unsupported. Devices run Windows 7 will be in the same boat as devices running Windows XP come January 2020.
>There's no reason to update devices that were never designed to change
Unless part of their functionality is to withstand attack from attackers whose knowledge is constantly growing. And pickable locks are the only thing on your list that qualifies. And as far as that goes...
We have pickable locks because an unpickable lock is apparently impossible, at least while being remotely easy to use. And locks evolved a LOT before they reached their current state - which are secure enough to deter crimes of opportunity (i.e. they keep an honest man honest). It takes hours of practice, or moderately expensive purpose-built tools, to get good enough to pick an average modern lock - too much effort for pretty much anyone without premeditated criminal intent.
Invincibility is too expensive, even where it's possible. Security is all about lowering your risk by increasing the cost and risk to the attacker. And when any idiot who picks up the electronic version of a free set of decade old automated lockpicks from the corner website can walk into your house without even trying, take what they want, and trash the place, secure in the knowledge that they'll almost certainly never be caught - then you don't have any security worth speaking of.
--- Most topics have many sides worth arguing, allow me to take one opposite you.
How often do you update your router? If your up time is over 60 days you are missing updates and are insecure.
I don't know any home/small business router company (TP-Link, Linksys, Netgear, ...) updating routers every 60 days. More like 1-2 times per year, for 1-2 years. And then nothing.
Perhaps you should look into Asus, which often updates at least quarterly, and often monthly:
* https://www.asus.com/Networking/RTAC68U/HelpDesk_BIOS/
* https://www.asus.com/microsite/2014/networks/routerfirmware_update/
And has been doing it for 4+ year-old products. Plus there is third-party code that leverages the GPL stuff that Asus releases:
* https://asuswrt.lostrealm.ca
* https://github.com/RMerl/asuswrt-merlin.ng
They buy it because it's better. It's better than Windows Phone (the first, second, theirs, and fourth attempts), it's better than Symbian, it's better than everything else people have tried. Why is it better? Linux is or reason it's better. Even Microsoft is using more and more Linux now. Is that because Microsoft has a religious zealotry for Linux? No, it's because Linux is better. Better than eating their own dog food.
>> Legacy software forcing people into Windows nowadays.
> Yeah, more than a billion people.
Yeah, legacy software has a LOT of people (companies, really) still stuck on Windows. Your point is?
I was personally very upset when Motorola refused to provide me a software update for a device, designed for both long-term and short-term use!
It was an SN74LS139N Motorola Dual Decoder 2-4 Line Plastic TTL chip.
How dare they deny me software updates for this chip containing two inverters and four AND gates!
I don't give a damn that they designed it for embedded use, I should be able to update the software running on it!
Right?