#1 > Or disable a hospital.
The entire Ascension Healthcare system of hospitals (142 hospitals, 2600 total facilities) in on divert since 8 May because they had to switch back to paper records. Change Healthcare has lost $872M since it was attacked in February.
Maybe it's more like the pandemic: seems like nothing, unless it affects you.
https://en.wikipedia.org/wiki/Ascension_(healthcare_system)
https://www.wired.com/story/change-healthcare-admits-it-paid...
#2 > Does your stuff need computers working 5,000 kilometers away? [implying that's bad]
What if you live on the Gulf Coast, exposed to hurricanes? You want compute resources warm and ready far away from that region. After Katrina, the Tulane medical school was able to re-form quickly because the noteservice was running a bulletin board forum on a VM in Romania. Everything else was underwater.
#3 > This is the sound-powered phone
Have you used a sound-powered phone? I managed damage control in a ship. Sound powered phones barely works. And the coordination system to actually fight that fire requires radios and making overhead announcements that definitely depend on electrical power.
#4 > They tried to sort of renew this emergency telephone network
When the entire San Diego region lost power during rush hour for 4 hours in 2011, the cell phone system still worked. I was able to email documents to Tokyo from a car despite no traffic lights.
#5 > Because if the cable to the US is down
Sure, but there are a lot of disasters where the cables are fine. Graceful degradation is all about having widely distributed options. Lots of people have What. Signal is even better for people with more serious responsibilities, IMHO. And, friends, if you think IP networks are vulnerable, get yourself a starlink terminal and a HAM radio license.
The question is, what is the cost to secure? I've been in so many meetings where the cost of security is 10-15x the cost of a breach. It's horrifying.
It’s an industry-wide issue that permeates every level of the stack. And so yeah, individual companies trying to retrofit security onto a jenga tower of technology is going to have to spend a ridiculous amount of resources to have any kind of impact.
I don’t know what the answer is, but I too believe things won’t change until the day someone figures out how to push a “kill all humans” OTA update to all the self-driving cars on some random Tuesday afternoon.
Even in that case I’m pessimistic that any action will happen. People will go on TV and say grave things, hearings will be held. Fingers will be pointed. Task Forces will kick off. Reports will be written. Bureaucrats will have stern conversations with bureaucrats. Politicians will say: we must this and we shall that. IT companies will sell their “solutions”. But no actual action will happen. It will be all talk and commerce but no actual hands unplugging and plugging in cables. We have completely lost the societal will to actually do anything besides generate words and reports.
When there is a real problem, people act upon it (assuming society is functional - otherwise the country simply dies). That’s why there is no better training for war than war itself. Ukraine has already unrooted all of the peace & love & no armament folklore in France, and even turned a lot of ecologists into pro-nuclear voters.
So yes, I wouldn’t be surprised if guarantees of offline mode (with regular drills) were passed into law for electric cars and everything cloudy, including IntelliJ.
> It’s an industry-wide issue that permeates every level of the stack.
Can you explain? I don't understand. Here's my take.
Let's start from the bottom of the stack. CPU has some good security protections. They have ways to ensure that boot code is signed. They have hardware protection for memory. They have memory encryption to isolate VMs. They have many amazing security technologies. I can run VM inaccessible by host.
Let's move to OS. Well, there's lot of security stuff in any OS. Process isolation, namespace isolation, encrypted storage.
Next level is container orchestrator which happens to be Kubernetes these days. Again, there's lots of security stuff there. Built-in and add-ons. Everything is authenticated with cryptography. Many ways to implement very granular secret sharing. Secret stuff is encrypted in-rest.
Next level is application framework. Can't tell for every framework, but all frameworks I've seen so far was quite security-cautious. They try to safeguard known security issues (like SQL injections), they make it easy to add security layers on top, and so on.
Nobody cared about security in 1984, I guess. That's not the case anymore. Everyone cares about security. May be there's still space for improvement.
The only people who don't care about security are end-users. They don't even know what security is. They don't care about their passwords. They don't care about sharing their access. They don't care to check domain before typing password.
Also some application developers don't care much about security, that I admit. But that's not the every level of the stack. That's the last level of the stack.
How can the user run some random application and not have it wipe out their OS? Can they plug in a random USB stick safely?
There's no real security in the OS part of the stack, but lots of security theater.
In my opinion, security (as well as Quality, and things like error handling, accessibility, and localization) is something that needs to be planned and implemented, from Day One.
Do a better job from the start, and the cost will drop like a stone.
People build walled garden security models because security imposes a pretty large operational cost on everything else.
Think of it as a “pegboard.” It has a bunch of holes to hook things onto. You make sure to brace it well, and use good masonite. That way, you may not know exactly what you’re going to hang on it, but you have a good infrastructure for it.
We are talking about war-like situations, and where one state actor has incentive to cause maximum harm to another. Exposing your infrastructure like this is unlike damage that can come from natural disaster. For example, disrupting the communications exactly before the attack. Similar issues (though through lower tech hacking) happened in 7th of October during the Hamas attack in Israel, where the over-reliance on advanced, complicated technology became a liability.
The stuff you describe make sense in normal, peaceful situations, where the cost of securing certain infrastructure can be higher than the cost of a power cut once. That has nothing to do with what the article really says, which is basically that infrastructure is currently not as secure from a potential hostile state attack. Also, in that case, a hostile state actor can combine attacks that together cause more damage than the sum of the attacks independently.
[0] - https://www.axios.com/2023/10/10/hackers-ddos-israel-hamas-c...
The whole idea of the internet (and even some of our infra, like suburbs or highways/rail) is that there's no one single point of failure. Like designed-to-survive-nuclear-war redundant.
Definitely incorporate the most advanced tech you can for when things are going smoothly to get that efficency gain, but there's a reason all branches of the military (that I'm aware of) still train and test their aptitude using paper maps and trig instead of relying 100% on GPS and electronic devices.
Sure, the routing algorithms can quickly adapt to changes in network topology, but they assume infinite bandwidth, which hasn't been the case since a long time now.
In other words, if a couple of important pipes disappear between tier1 peers, alternate routes will certainly have trouble handling all the new traffic, which would make everything grind to a halt, and will only be solved by pissed network admins null-routing that additional load.
I know it's controversial in the context of net neutrality but personally I'd be okay with traffic shaping/prioritization for critical infra in cases such as this. Keep the power plants, emergency services, military, government, transit running over intsagram and netflix when things come down to it.
1) There is a dedicated microwave link between Vandenberg Space Force Base and Edwards Air Force Base. Mil. owned and operated solely for their own use.
2) The US Federal government decided to build a standardized communications network for government/first responders/etc. This is FirstNet. They farmed the build-out to AT&T and gave them 20 MHz of bandwidth (Band 14) but it runs over their standard wireless infrastructure and network but FirstNet traffic gets prioritized.
https://en.m.wikipedia.org/wiki/First_Responder_Network_Auth...
I think all three are physically seperated from the commercial internet and each other but don’t quote me on it.
Uncle Sam doesn’t mess around…
The reality of course is that the internet has turned into a fragile, centralized system of complication that rests on single failure points like Cloudflare, AWS, and Chrome. The internet as envisioned by DARPA would have survived to be used by cockroaches, the internet today would not survive.
It's just how it's been used by the majority for the last couple of decades that's fragile.
Thinking about this though it’s really the big tech companies manufacturing “the latest thing” to be tossed in the bin after a year. Dollars over longevity. Then they become “no longer maintained.” Could we STILL use a 3g network? Or is there a simpler, slow network that should be good enough barring our pointless desire for cat videos?
And some folks wonder why companies still use floppy disks on air-gapped infrastructure. Because it fucking works don’t litter it with complexity to modernize.
Now… the situation with skills to manage infrastructure? Now that the whole AI thing is happening? The internet is going to be fucked people. It’s time to go analog.
That’s essentially www, not Internet. You don’t need any of those to communicate (reliably) between two hosts.
On my most recent trip there - I was amazed at how despite being routinely hit by missiles, their train systems “on time” status is better than British or even German trains.
This is only possible because their railway workers have balls of steel and go out to repair damage fast, and sometimes get hit in follow up strikes.
Same with energy workers - they go out and repair stuff during air alarms, in the immediate aftermath of strikes they perform damage control and mitigations.
That said, paper based redundancies do exist as a massive ransomware attack is similar in impact to a multiweek power outage.
Around me, cell towers have 3-5 hours of battery when utility power is out. If your outage had gone on much longer, you would likely have seen cell towers start dropping out.
Of course, my area also has some other nasty SPoFs. A couple years ago, a telco cable was severed and DSL for everyone was out and at least some of the cell towers were live, but no service. A few weeks ago, the cableco had its wires severed, and cable tv and internet was offline, and so were some cell towers. IIRC, for the telco one t-mobile worked and verizon didn't, and for the cableco t-mobile didn't work and verizon did. Not sure about at&t.
You have to understand that this article was written by an European technologist worrying about a war situation. Sure, you can make a counter-point, but your counter-example is very different in many aspects: nature of the threat, jurisdictions involved, orgs involved, etc.
> A failsafe on-board multi-sensors localisation unit consisting of a navigation core (IMU, tachometer, etc.) brought in reference using GNSS, track map and a minimal number of reference points
From https://www.clug2.eu/objectives/
> to complement the existing European Train Control System (ETCS) odometry system by using GNSS to enable absolute safe train positioning whilst also transforming today’s train localisation by demonstrating a GNSS based multi-sensor fusion architecture.
Okay, so as I expected they want to add GNSS as an additional sensor input. That is useful because without it train odometry is purely relative and the train doesn't know where it is until is reads the first balise. The plan doesn't seem to be to remove all other sensors. Denial of GNSS would then mean that start-of-mission is about as tedious as it is today and odometry accuracy might be reduced. Depending on the number of balises on the track that lowers the capacity of the track a little but is far from catastrophic.
Second, in a war-like scenario, do we really want to have our critical infrastructures unable to operate at full capacity?
Railroads can now outsource train control. Wabtec's "Wabtec Cloud Positive Train Control Communication Solution" - "A complete turnkey hosted office solution for I-ETMS-based Positive Train Control (PTC) systems"[1] (Wabtec used to be Westinghouse Air Brake.)
Wabtec has had break-ins, but claims they only involved employee info, not control systems.[2]
[1] https://www.wabteccorp.com/digital-intelligence/signaling-an...
[2] https://industrialcyber.co/ransomware/wabtec-suffers-data-br...
So, my hat off to you, Internet stranger.
They should have lots and lots of local transmitters.
The absolute last resort for trains is semaphores and mutexes based on physical tokens. Those concepts came from there, and were still used sometimes to this day. Doesn't sound high tech, but it works.
Critical infrastructures should not depend from things that are located in space or on the other side of the planet. These are one of those things were market logic should be anticipated with regulations (we can’t wait for the next Titanic). Another point touched by the article.