Workers in such a facility should not have their general purpose phones and computers with them.
This is not a panacea - there are conceivable attack vectors for non networked equipment - but it raises the bar significantly and should be the bare minimum of acceptability for designing and running these plants.
Ok.
> The equipment and tools should not even have a network stack
How do you imagine sensors and actuators will be operated instead ? People running about everywhere with handheld radios, reading gauges and turning valves ?
I know a guy who develops simulators for refinery operation. He explained to me how each refinery is a unique setup, cold-starting one takes a couple of days of non-stop valve twiddling with eyes on gauges and has plenty of ways to end in a fiery inferno if you do it wrong. Computerized control systems are not optional.
Don't bring a thermal overrun detection to the central computer and have it shutdown a burner. Do this locally
At least we've got a set of verified, battle-tested, multi-platform TCP/IP stacks. More importantly, they have been built with generalisation in mind, meaning that a single implementation would encapsulate all of the use-cases.
But instead we've got those half-baked implementations of "oh my god, it takes me only a few hours to establish a radio link" protocols just increasing the attack surface.
You're speaking of a semi-technical solution to a problem that is directly caused by the race to the bottom in the manufacturing industry (they probably have installed that equipment a few decades ago, just before they fired the last of the employees).
That's not even an argument. You can put a buffer overflow in anything. The key thing is this: the larger your code is, the more likely you are to have one. Using serial lines isn't going to make your code larger if you're just taking sensor readings or something. Far from it.
> At least we've got a set of verified, battle-tested, multi-platform TCP/IP stacks.
Think about this in terms of Linux or any other Unix or bigger RTOS, if that's most straightforward: in the kernel, there's much, much less code needed to make a UART appear as a device node than there is to make an entire TCP/IP stack and network PHY work. In the kernel it's all "battle-tested" code, none of it more so than the UART drivers. In terms of software quality, relying on the serial port and not using the TCP/IP stack means you're exposed to fewer potential errors simply because there's much, much less code being used.
(edit: the same basic idea applies if you're talking about something more sophisticated than a UART, like a CAN bus, which is still way less code than a TCP/IP stack)
> But instead we've got those half-baked implementations of "oh my god, it takes me only a few hours to establish a radio link" protocols just increasing the attack surface.
In terms of system engineering, you're exposed to fewer potential attack vectors if someone needs to plug into your system via a serial line versus accessing the system from anywhere you're connected to using the highly fragile resource of TCP/IP routing, but even connecting via a short-range radio link need not be much more complicated than that. Use a radio modem that connects via serial if you're using a real computer. Similarly, if it's a tiny embedded thing the radio is probably exposed via a serial interface.
Adding TCP/IP just adds code and attack vectors. That's fine if you truly need it...
They were before computerized control systems existed.
People have been cracking petroleum molecules since 1891.
I don't have to imagine it - I can just look back to how it was done before the widespread adoption of general purpose computers and their networks.
The systems, and their controls, should be as dumb as possible and should fail in as boring manner possible.
That merely hardens the target, but does not protect it. Stuxnet was delivered via USB to an Iran nuclear facility and had the exact same intent of triggering physical damage, which it apparently did successfully.
A decade ago, disconnecting from the Internet would have been considered security, but things have progresses and security needs to as well.
In my admonition against networked, general purpose computers, I purposely neglected to also mention that the plant engineers should not shoot themselves in the face, with an actual gun, upon starting their work shift.
I neglected to mention this in the same way that I neglected to mention that you should not have a PC running Windows anywhere near your critical infrastructure.
The failures that allowed stuxnet to occur were NOT "casualties on the battlefield of modern cyber warfare" - they were the clownish actions of idiots who got owned by autorun.inf[1] (among other things).
[1] http://www.eweek.com/security/how-stuxnet-malware-used-autor...
I agree with you that they should not be connected, but you can have things disconnected and still be read-only. IE, one way optical feeds.
Otherwise, you'll always have some goon plugging in a wi-fi router, rigging it to a sat phone with a modem, all just to play quake with their friends, while they idle on some crane barge, or wherever their stuck, bored on some hurry-up-and-wait project plan.
Even then, you'll still have wonder about silicon bugs, or tempest and row-hammer style attacks, for anything software ever touches, when a given industry fails to identify such threats themselves.
If you come up with a different network, then the attackers will just switch to that network.
Remember, the internet used to be lots of different networks. I'm old enough to remember bang paths, BITNET, and e-mail taking a week or more to make it across the Atlantic. Even back then, there were vulnerabilities, and even cross-network vulnerabilities.
And a private satellite is a terrible idea for "security." In a previous life I used to have to operate a satellite uplink, and I can tell you that replacing a transponder's intended content with your rogue content is really quite easy if you just put your mind to it. (It's happened before.)
Any reading material behind this? Really interesting, would like to read up on what caused such a delay
A large part of the delay is that messages were transmitted in a store-and-forward scheme (often via uucp). And most machines didn't send messages more than once a day because connectivity was expensive and measured in dollars per minute. And when they did connect to the next machine, it was usually not a very long hop. Sometimes one part of a campus to another. Or to a computer in the next town.
I really don't know exactly how the messages crossed oceans. Satellite transmission would have been unthinkable. My guess is that eventually they hit some big east coast computing center like MIT or BBN and went via undersea cable, but that's speculation.
I ran a node of one of the pre-intenet networks. Like most of the other nodes, it was connected via 150 or 300 baud dialup modem. Later there were a few 1200's, but they were rare.
My node was important because it was oddly located so that it could span two states and two area codes without incurring toll charges. That made it very busy, so the early morning (2am) message transfer sometimes took a couple of hours.
I tried to write a couple of articles about the old American dialup networks on Wikipedia once, but someone in another country deleted them saying they didn't exist because he'd never heard of it and if there was no web site to link to as reference it didn't happen. I stopped contributing to Wikipedia after that.
Yes, I get that electronic signaling is electronic signaling, and none of it is actually different, at the transmission layer. It's just more DSP and more fast fourier transforms under the hood.
Yes, technical barriers can be eroded with adapters and facades, but it's an added cost to attack, and reduces detective work in that you have to know someone to jump the learning curve and enter the attack envelope. That means detective work can happen within a smaller social graph, and that alone becomes a deterent from sharing information, because everything becomes need-to-know, and insider awareness is a give away for inside jobs.
It's also easier to stamp out, and ostracize insiders, if they have loose lips or have a tendency to lend and give away the car keys.
Not everything needs to be as cheesy as Encryption DRM for optical movie disks and video games. For critical infrastructure safety is important enough to warrant independent military-grade safe guards.
Do military protocols fail? Yes, we have the enigma machines sitting in enough museums to prove it.
This in not an XKCD "too many standards can be solved with one more standard" concept. Isolation and specialization can be effective defense concepts.
Thank you - appreciated.
Many people (unwittingly ?) argue against defense-in-depth because they look at individual layers of the defense and declare them inadequate. They are always correct.
What they are missing is that nobody ever proposed only using (port knocking, or stack obfuscation, or fake login banners, or whatever). They are always additive layers of security on top of the existing set of best practices.