HN Simulatornew | past | comments | lists | submit | m4x's commentslogin

I'm curious for some elaboration on the proposed problem - are you suggesting that "Take a paper with dimensions x, fold as follows, then describe the result" is something an LLM couldn't currently answer correctly?

My initial feeling is that this is something an LLM would be quite capable of, and the gap is likely a tooling and specification issue (i.e. you should first develop a method for simulating origami folds, then build on that)

A non-savant human certainly couldn't answer the question as posed, either. They would answer it by actually following the steps and seeing what happened.

Have I misunderstood what you're proposing?


Look up "Old Dragon" by Travis Nolan. I chose it because it's an intermediate-level model with, as far as I can tell, no published step-by-step fold sequence or video tutorial. There is only the (monochromatic) crease pattern, i.e. what the model looks like when completely unfolded. I am asking for a fold sequence given the crease pattern.

Given that description, it might sound about as hard as reassembling a shredded document. But the truth is that it doesn't take a savant when you're familiar with the most common patterns, and that's the reason why folders online generally share their models this way. Yet legit academic research on doing this with computers even for simple models is primitive (see Creasy/ORIZU for the closest prior attempt, or more recently OrigamiBench documenting essentially "LLMs suck at this, even with tooling"; if you dig around you'll find some interesting work by Lang suggesting that the admittedly quite general technique of tree theory that he has algorithmically pioneered since the 90s can be driven by AI, but that falls far short of a general understanding of origami).

Anyway, obvious caveats being: I am under no delusion that AI couldn't become superhuman at this given a clever enough vectorization of the problem amenable to deep learning, and/or a huge corpus of richly labeled training data, but I stand by my claim this is a well-defined mathematical problem type that current AI cannot solve just by asking it, while a skilled human can.


Surely there are two separate transgressions: one being the "theft" of legally protected knowledge without following legal processes, and the other being the commodification of creative work that people still (arguably quite rightly) put on a pedestal.

I think people would generally be happy with the IP theft part if the result was actually beneficial to society. People are not happy because in addition to the IP theft, creative work and human labour in general has been further devalued.


> the other being the commodification of creative work that people still (arguably quite rightly) put on a pedestal.

But if we take a strong position that information should be free, the kind of commodification of creative work we're talking about can't really be avoided, and doesn't seem to violate the spirit of that ideology.

The reality is that a pure "all information should be free" position goes well beyond what most people are comfortable with. If current IP laws were eliminated, people would want something else to replace them, and information still wouldn't be truly "free".

Notice how e.g. GNU licenses are entirely dependent on copyright law to have any force. This is sometimes spun as being a clever way to use the system against itself, but if copyright law didn't exist, several restrictions that GNU licenses impose would no longer be enforceable, and we'd suddenly start hearing from Richard Stallman again.


I am in agreement, although there's a bit more complexity to this. A substantial amount of research is funded by commercial partners, and even when research is primarily funded by government the intention is that it moves to benefit commercial partners ASAP. If you publish this research freely, then commercial partners typically lose the competitive advantage they paid (or were hoping) for. The IP and commercial teams at many research orgs would therefore argue that the public actually sees more benefit from not publishing.

While I can see the logic behind this, I would prefer to see knowledge published rather than locked away for commercial benefit. This would require a fundamental change to our economic system, however, in addition to making research "more open".


> The laptop class doesn't always grok that the shift-work labor class ...

I think it's more than this. Industrial control system upgrades are serious business and (at least if you're doing them properly) require a lot of planning and consideration. Particularly when the subsystem being replaced is decades old already.

I think the overwhelming majority of people (including on hn) have no understanding of how complex industrial systems are - even as-built, but particularly after decades of evolution. Doing proper engineering in these spaces is expensive.

You can reasonably blame supermarket chains for being insecure, but we certainly aren't currently enabling them (or anybody else) when it comes to fixing these issues.


Is it more expensive to keep your systems upgraded and maintained or to have all your product get destroyed one arbitrary day and then have to pay to do the upgrade anyways so that the arbitrary property destruction doesn't keep happening?

In before "but insurance"; yes insurance can in theory mitigate some of this, but I feel like people on HN have no understanding what it's like to deal with a multi-million dollar insurance claim that the insurance company 100% does not want to pay out. The legal battles over such things take years and are on their own incredibly expensive.


The thing is that many IoT devices aren't simply updated. The industry doesn't work like software development countries.

They usually think in projects. Some project manager, some funds and some workers are assigned to a specific project. And than this project is followed through. Eventually the project is done (finished or dismissed). And all the allocation shifts to the next project.

This is e.g. the reason why most of the Android phones have so bad security. Even Qualcomm does the Android stuff on their chips only project wise. Is the project done, then you'll get an Linux kernel from 2021 with this specific device. There will never be an update, even if this outdated Linux kernel has 20 known CVEs. The reason is simple: no PM allocated, no funds allocated, no workers allocated.

And such project-thinking also exists for other hardware chips, e.g. network or WIFI chips. Lots of development to get the thing "ready to market" --- which isn't necessarily to get it good --- and then almost nothing.

And this is the hardware the industry automation uses. Except that there some intermediaries in between, e.g. Kontron that make industry-grade boards. But even at that level you have the project structure. These boards are then used by the manufacturers of the industry automation which, who would have thought it, also mostly think in projects.

And this is the reason why the S in IoT stands for security.

People at various levels would need to switch to a process mode, away from project mode. And people at various management levels need find a way on how to sustain this financially.


> Is it more expensive to keep your systems upgraded and maintained or to have all your product get destroyed one arbitrary day and then have to pay to do the upgrade anyways so that the arbitrary property destruction doesn't keep happening?

It happened once. Did you do the maths here? Can you tell us which one was Kore expensive and what percentage chance there was of this hack happening?


Someone did the math. Cost of continual maintenance, versus cost of product loss. They determined it wasnt worth the maintenance cost, and accepted that risk. And that risk manifested.

This happens all the time. Unfortunately, in my experience, most of the risk acceptance happens without even really believing in the cost of the risk, and people just accept everything. it is highly likely someone like that accepted this.


And the people that made the decision largely have moved on and cashed their bonuses from saving all that money. It's now someone else's problem.

Management is never held accountable.


> Management is never held accountable.

By whom and for what?

The problem with security maximalist[0] ideology is that it does not accept reality for what it is: shit will happen. The correct way to handle risks is to mitigate them to some extent, and then insure (or self-insure) against the risk that remains[1].

Then, when the risk actually materializes and mitigations prove insufficient, you write off the damage and carry on with your day - and yes, perhaps even reward the executives who set up mitigations and insurance that allowed the business operations to continue undisturbed.

Only then, mitigations MAY be upgraded and insurance premiums adjusted, as an event happening is an important new evidence in risk estimate for it happening again.

That's how adults in the room handle security.

It's only the general public and cybersecurity people who see random events happening and decry that somehow it affects them and Somebody Needs To Be Accountable, and Someone Must Do Something About It.

--

[0] - Or should I say, cybersecurity maximalist, because in every other field except software, this is well understood.

[1] - Talking about regular security, that applies to 99.99%+ of businesses; the risk to life and limb is a special case with extra considerations, though it fundamentally doesn't change much about the framework.


The polite term is 'self-insurance', and it's often the right move.


That class of risk was just theoretical for decades, so the price on it was basically 0.

Now that we see large scale incidents will there be an actual money value associated with upgrading systems and security for supermarket appliances.


I think it's fair that the price up until now has been mostly zero. On the other hand you can 10+ years of examples on HN of people saying these types of threats with IoT devices and internet connected industrial hardware are not theoretical and it's shocking they weren't being exploited more often.


I can find 10+ years of examples on HN of people saying all kinds of stuff, and it's shocking how outlandish that so much of it is.


The thing I can‘t wrap my head around: why does a freezer need internet access at all? My freezer at home works just fine without. I bought it 20 years ago. Worst case in the supermarket: a freezer dies on Saturday night, you throw away the goods on Monday morning and call someone capable to fix it. Lowest TCO solution.


My random guess: freezer need temperature audits and franchise supermarkets have to prove they're properly managing the food. If the central monitoring service can also alert/dispatch staff within minutes to deal with issues, it becomes a pretty attractive proposition.

Something that might not be obvious, but temporary dips in temperatures also need to be monitored as it's enough to spoil the food. Worse case scenario isn't to have to throw 50~100 kilos of food, it's to have spoiled food still looking fine and consumers ending at the ER, with the whole franchise name on the news.

> My freezer at home

That can't be your frame of reference for a business entity part of a franchise.


There are better (and offline) solutions to detect temperature dips in the freezer (e.g. a thermograph, or cheaper, solutions that change color). And as a former trucker delivering frozen and refrigerated goods I can tell you the most dangerous part of the journey to the customer is not in the freezer of the POS. That being said, I admit I don’t know the legal requirements today. But I doubt someone will be sent out at night to salvage a few pounds of frozen spinach.


On better solutions managed locally, yes of course. Small shop's also deal with the same requirements without keeping their stuff online.

> someone will be sent out at night

We're talking about France, so that's an even lower chance of happening for a freezer's worth of goods.

The two upsides I'd see are the ability to throw away the food next day before opening the shop, and the franchise being able to force that procedure even if the local shop would have chosen otherwise.

Local tenant are usually a lot more lenient than their franchise manager (for good reason usually, but there's always that tension)


and the price of everything will go up as well to pay for it. Your $2 bag of flour becomes $4 because of security costs


this is sad. its like tragedy of the commons. you could have lived in a village without thieves and you dont need a lock. but as soon as there is one thieves, everyone in the village now need to have locks. you could have all your servers without security in a village with social trust but the moment you have evil actor everyone must now have security expenses that do not add functionality just to fight that single pathogen.

so the expensea are related not to amount of adversaries but to amount of things you need to protect. because its enough to have one bad guy to inpose expenses on whole village.


It will shortly become a national security issue.

Russia is feeling the pain of having their warehouses blown up.

There will be tons of budget for having things ‘blow up’ themselves.


That is a dangerous game, especially when affecting food supply. Widespread cyber attacks on critical infrastructure are deemed grounds for nuclear response in most national doctrines. This obviously is way out of scale, but concerted widespread efforts by decentralized groups could end up triggering ww3 by accident. Not when they try to play a game of chess, but when they shut down a water treatment plant, and some other unrelated group shuts down gas pumps, and suddenly there are millions of people with no fresh water and no fuel to get it to them...


Yup. Plenty of proxies to attack through first, of course.

Cold War, part Deux - the cybering?


It's psychology. One option is a guaranteed upgrade bill that's due right now. The other one is a speculative what–if that may never happen.

And that's even assuming anyone who runs these sorts of systems have "hacked to turn on defrost mode" on their bingo cards at all, which probably isn't the case.

On top of that, upgrades carry risks of their own!


It was „what–if that may never happen” before agentic AI.

Nowadays you can be pretty certain it will be exploited.

It is not just run faster than slowest guy to get away from the bear, now it is dozens of bears per runner and bears are stupid enough to chew on anything but also smart enough once they bite something tasty that works will just continue biting that.


In my recent experience (last 3 months) with a cyber claim getting insurance to pay out was actually pretty straightforward. It was about a $250K USD loss on a phishing attempt (someone hijacked and rerouted payment details for a wire)


But you don't need to replace entire system to massively reduce surface area for the attack. Just do what was done for decades, and put the vulnerable devices between VPN and/or additional authorisation layer.

It's far less work than touching anything on controller side (at most changing some network settings) and requires attacker to at the very least penetrate your network perimeter and snoop around to find it rather than "run nmap on a controller's default port over country's IP range"


don't connect these to the internet that's how it was done


That's effectively what "require a VPN" means. You can simply block any connection that doesn't come through the VPN.


a vpn is not as safe as you think it is.


An enterprise grade VPN is pretty damn safe. A government could probably get in but then a government probably isn't going to bother hacking your fridges to defrost.


Well, unless you buy one from Fortinet.


> I think the overwhelming majority of people (including on hn) have no understanding of how complex industrial systems are - even as-built, but particularly after decades of evolution

I may not have experience, but it is standard security practices that monitoring and control must live in separate channels for security. This is not a "planning and consideration" concern, this is conceptual design of systems.


why would I connect my appliances to the internet in the first place?


Just an hypothesis:

A supermarket may have 30 freezers. You don't want them to beep loudly if someone forgot to close the door. Instead you connect them to a central monitor. A local employee can see on the monitor which freezer has a problem and go see the issue.

The monitoring is as well a proof that the cold chain was kept.

This network was local to the market. Then one day it was found better to collect these data to a modern dashboard, and so the local network ends on corporate network.


> You don't want them to beep loudly if someone forgot to close the door.

Why not? It works as an alarm, is simple in design and requires no external input.

They need to be checked 3x per day anyway for temperature so someone will notice eventually. And checking temperature must be product core temp so no remote monitoring. Unless haccp changed massively over the last 20 years to "we don't care anymore".


This is remote management, not just monitoring. Say you want your fridges to integrate with weather predictions, or prepare for a change of stock in a specific freezer across 100 sites. Features are appealing, but hopefully LLMs can make these cowboys finally secure implementations that they have never been forced to before.


An audible alarm encourages people to tamper with it when it breaks and starts misfiring. Sure, they should put in a maintenance request, but they frequently take things into their own hands, even if maintenance is good about these things. And if the audible alarm just dies and there is no central management, how would you know? Stores don't really want to have to pay people to stand around with a stopwatch to determine if a time-based audible alarm still functions properly, and they also don't want to scare customers when things go wrong or when the alarm correctly goes off for a door just slightly ajar (and thus it wouldn't be immediately obvious which one needed to be closed).

With centralized monitoring, this all becomes much easier, and you can inspect a whole row of freezers in less than a minute. You can even do it remotely, just watching when customers open and close the doors on a security camera and checking that the control system registered it correctly.

See also: dump trucks that have alarms if you start to move while the bed is up. It's a great safety feature on the highway, but it's annoying on the job site if you can't just dump everything out while standing still, so it frequently gets disabled by the workers, especially if it starts to misfire. This is how you end up with dump trucks going highway speeds with their beds up taking out overhead signage and slamming into bridges. The stakes are much higher, yet workers would still rather rip out the alarm.


How do you measure core temperature of, say, a block of frozen fish without destroying the packaging and contaminating the good? Do they throw it away after measurements?


Apparently the freezers needed to coordinate defrosting. Evens out grid load or whatever. Making everything Internet controllable is one way to achieve that.


  > it is standard security practices that monitoring and control must live in separate channels for security.
Was that the case 30 years ago when this equipment was designed, built, and installed?


I started by saying that I don't have experience... so I will have to research that. Or you could save yourself some hours and research it?


I work in SCADA development, and believe me, there's a reason the industry moves slowly.


This slowness also means it sometimes has a hard time understand new risks and a lot of industrial systems are not operated by people who understand those risks and do…very stupid things.


100%; I've worked on many a project that was implemented without proper concern for cybersecurity. It seems like this hack in France was triggered via a missed update (according to the manufacturer of the system, who I am hesitant to believe tbh). Security updates aren't always done automatically, and if a system's implemented by a consultant and the owners of said system don't know how to operate it, then you wind up with a bunch of spoiled food across however many supermarkets.

The best systems are implemented either in-house or by a team of consultants working hand-in-glove with the in-house IT or ops teams, then maintained in-house. That's been my experience. Unfortunately your average municipal wastewater plant is probably run on decades-old systems by the same IT dept. that fixes the city council's printer.


This comment could be a lot more interesting, useful and worth typing than it is. :(


Sorry, it was a busy day. I was agreeing with everyone else here. The move fast and break things attitude is one we can't afford when building control systems for critical infrastructure. I used to work in big tech, and if I pushed an update that didn't work or wasn't tested properly, the worst thing that could happen was we revert the change and there's a minor inconvenience to a user. If Facebook goes down for an hour, the company loses money, some advertisers get mad, and their users get frustrated. If an industrial Siemens SCADA operation goes down for an hour, the best case is the operation (be it a factory, airport, wastewater plant, etc.) shuts down for an hour. The worst case is people get sick or die.

It sounds dramatic but every update or change to the system rocks the boat and exposes it to a whole new tranche of dangers. So we spend most of our time observing. There's no such thing as a casual QoL update in SCADA.


What I don’t get is, if it’s so critical, why wasn’t the bare minimum done to isolate the hardware from the Internet At Large?


I think for the big players, it often is, my impression is the more systemic risk is smaller players (like very local utilities) these are sometimes run by very unserious people.

Most SCADA systems date from the 1970s and a lot of them don’t really think about network security or even have a concept of it. So, they don’t think about the consequences of connecting it to the Internet for remote access. It’s just a feature they want.


> There's no such thing as a casual QoL update in SCADA.

> So, they don’t think about the consequences of connecting it to the Internet for remote access. It’s just a feature they want.

This is the disconnect that's so confusing.


It’s different sides of the house.

Operators understand how important having them connect to PLCs and various bits because failure can mean you can take out parts of whatever it’s controlling, the power grid, oil pipelines, etc. So, you need to be very aware of what you are doing when you do these upgrades. There’s a lot of process to ensure you don’t break things.

But, you tell them about network security and or having to have Modbus messages wrapped in TLS and some of them will tell you to shut up and get out of their way. These systems have literally been just ported from the equivalent of a PDP/11 to Linux. So, not the most robust things when you come from a network security perspective. They really are 1970s systems and it’s really annoying, but I don’t think there’s any money in really doing anything about it and a lot of people make good money consulting around for these things.


I don't work with many of the smaller players, but you're right. They're far more likely to not have their own teams building and managing these systems, so they're susceptible to the Lyle Lanleys of the industry who come in and show them how they can control everything from their phone. They get sold off-the-shelf products and told that they're plug-and-play, but they're usually not.


> believe me, there's a reason the industry moves slowly.

No profit in security/dependency upgrades. Propriety SCADA protocols to enforce full stack manufacturer lock in.


That too ¯\_(ツ)_/¯ then you end up with a bunch of civil engineers who build a monitoring and control platform in Power BI and expose everything to the open internet, which then gets exploited, and they wind up working on a Siemens product anyway (I don't work for Siemens, I do custom development). It's really frustrating to watch.


Go on.


Plus the lifecycle on these systems may be decades not weeks. At some point it needs to be segmented physically (more cost).

In 2017 I was hired at a place to support their ERP system and regular SysAdmin work. My first month was the last patch for their 10 year old ERP system. That was key software running the entire company and it was deployed before I finished middle school. The floor was run of scanners using WinME terminals that lost config on power loss.

That business is still profitable. I did not stay long.


The fix is taking them off the internet, use DOS, and require manual insertion of a floppy to make updates.


Use CP/M. It's the only way to be sure.


I get the first and last point, but how does DOS help?

You can use a modern operating system just fine.


I ascertain that they were using hyperbole as a form of humor.


DOS ensures nobody will be have an easy time reversing the decision to keep them off the network.


I'm fairly sure it's trivial to get Claude to vibecode you some little program that runs in DOS and allows you to connect the computer to the network.


Sure, it was called Novell. There was also AppleShare PC, which connected you to a Mac network. It was how I laser printed my dissertation. ;-)


And that was presumably before Claude made it easy. :)


Assume the computer does not have a modem of any sort!


If you assume away the hardware, then even Linux can't talk to the network.


Yeah the margins are tiny because private equity has spent the better part of the century destroying American dynamism and outsourcing all externalities including getting hacked to the people.


This is only partly true, historically before private equity was a thing grocery was always a razor thin margin business that operated on volume. PE just made it worse.


My in-laws had a family grocery store back in the 60s and 70s. Privately owned and operated, no way they'd have been paying any attention to security concerns like this.

Their only chance would have been if appliance companies wouldn't be dumb enough to put a simple refrigerator on the public internet in the first place.


My appliances will never be networked, for this very reason. Risk to reward is just not there.


I have an oven, washer/dryer, garage door opener, and lights that all want internet access. They don’t get it.


This is happening in France.


As the IT Director at an independent grocer that's currently overhauling it's HVAC system (super fun for me to see Danfoss in the news this way) let me assure you that no one working in this space day-to-day has much of a clue wtf they're doing. When I inherited our systems, all the controllers were statics set within the subnet's DHCP lease pool which was shared with the office network at the time because, and I quote, previous IT guy thought VLANs were scary. Every couple of weeks a new DHCP lease would be assigned that would overlap with a microcontroller's static and kick the communications to something like the main AC governor off the network.

The guys deploying the new system ($2 million for a single location) are having HVAC schlubs terminate CAT5E and leave the gateway controller just hanging, unmounted to the wall or anything. Lo and behold, the RJ45 port has been broken and you have to send me a new gateway.

All the replacement controllers (for a system that is supposed to be better than Danfoss) are being set to 192.168.0.x. Each individual device, based on the "technician" descriptions are running individual DHCP servers inside them. And they just deploy them, no asking IT about approval or correct network config, just deploy and then make it our problem on a random day when they realize nothing is talking to each other.

They are just now in the late 2020s realizing what a VPS with an API is - truly, nothing to worry about with baby's first public IP address when baby doesn't understand what the implications of multiple DHCP servers on a single network might be.

The people "designing" and deploying these systems are HVAC guys with a little bit of software and networking knowledge. The networking (and security lmao) are the lowest priority part of what they're doing, they're more concerned with crimping pipe and moving onto the next project.

The people doing the project management and buying decisions are likely not IT at most operations, but the Facilities guys, and they'll slap a microcontroller anywhere until IT catches them or tells them they can't do what they want as easily as they'd like. There are no security reviews before deployment because the people heading the project don't even know what security is if it isn't petty theft and they don't want to hear that it will take longer than the contractor is promising them.

For additional context here, given everyone talking about the tightness of margins in retail grocery, until 6 months ago our network backbone switches in an environment where we need true on=prem servers to run operations, were still 10/100 Cisco Catalyst 2000s. Previous generations had fucked up the networking so hard with such little documentation and wacky misunderstandings that it took me several years to unfuck bits and pieces slowly and then schedule an actual maintenance window (because those mean potential lost sales, can't do that!) to cutover the rest to a more reasonable series of 1Gbps switches. We do mid 8 figures in a single location and I had to go through an entire song and dance to get budget approval for scheduling the 4 hours of OT for my sysadmin to do a maintenance window with me at midnight. Granted, we post mid 5 figures in actual net income most years, so yeah, margins are incredibly tight for smaller players and those of us actually trying to do a good job in any of the different sectors of the business are constantly trying to fix messes that wouldn't exist if there'd been better understanding or honesty about the long term costs of short term savings.

In any case, I'm sure there are more business segments out there like grocery HVAC where the networking is a simple means-to-an-end and the people deploying these networks barely understand the depths of the waters they're swimming in. There is absolutely a need to coordinate all the constituent subsystems to keep inventory stable until it sells, but yeah, if I'd even have been in the room during decision time with the contractors I'd have shot down any plan with external access without very precise deployment specs and plans. But I wasn't, I'm just the guy finding out some just certified HVAC chucklefuck is deploying DHCP bombs on my network 4 months into their contract and thinking they're doing me a favor. Without a doubt, a sufficiently informed and motivated attacker could knock out many SMBs that don't have a suitably knowledgeable IT department or god forbid an MSP charging them extortion rates for little real value.

Really love that I'm finding out our current HVAC system is getting exploited in France and at US Army bases on the first day of my vacation. Honestly think my only winning move here is to either buy a completely separate ISP ingress for the HVAC system or set a new subnet with only enough address space for the handful of microcontrollers needed for this new system. Time to start pushing on when the Danfoss is being retired and with how much gusto.


I’m curious, which system are you ripping out, and what are you replacing it with? I sell and run HVAC controls installs and do business with JCI, Siemens, Honeywell, Automated Logic, and Daikin. Sometimes we install Schneider Electric, Alerton, or Distech for independent dealers.


Great writing. Maybe you should write more about your experiences. I would gladly read more.


How does this actually work, in practice? If you have read access to sensors, then there's a non-air-gapped network between the control systems and the internet.

If your network isn't air gapped, then it's vulnerable, and attackers will side step the "read only" path to affect control system changes.


It's being called a data diode (diodes usually allows current in one direction only).

It can be implemented with a fiber optics cable with only the transmitter on the secure side, a laser+photo cell for real air gap, cutting the rx cable in a rs232 or having a microcontroller with an output pin connected to another controllers input pin. By handling it with hardware instead of software you get a very secure partition between the trusted and untrusted side.

It's used to monitor for example industrial systems and power plants.

Wikipedia calls it unidirectional network: https://en.wikipedia.org/wiki/Unidirectional_network


Yes, this is exactly the kind of approach I'm referring to when I say it's possible (though not trivial to implement), but it still doesn't prevent attacks. Air France 447 crashed with extraordinary loss of life because the pilots responded incorrectly to reported sensor readings. Any scenario where authorised staff are able to affect changes in a control system is vulnerable. The attack vectors just change to suit the target system.


AF447 wasn't a security breach. You're talking about an entire separate class of human-machine interaction problem that has little relevance to the lack of security around access to OT operational surfaces.


Not OP, but the point seemed clearly to be that if sensor input is wrong, then it can result in catastrophic results even when all control systems are normative. The inference then is if hackers can affect sensor readings then they can still effectively control a system - like putting an ice cube on a non-adjustable thermostat.


If hackers can "affect sensor readings" before they feed into the control system, sure. But that is not read only access to the sensor... The trivially simple example is a webcam streaming the thermostat output. The hacker can watch that stream all day - it's not gonna cause any issues with the refrigerator. Even if the hacker had control of the stream, worst they could do is disrupt any monitoring system that uses the stream (e.g. make it appear that there is a fake issue or mask a real one) - but that is significantly lower risk than having remote access to trigger a semipermanent defrost cycle that spoils all the food.


You'd expect the people reading the data to criticize it and not input any command that will compromise the operational safety.

Or, in other words, you'd expect the managers to be smart enough not to set all the fridges to 50°C, whatever data they read on their screens.


All fancier than a camera pointed at the thermostat readout I’m sure but is the laser & photocell anywhere close to similar in concept?


Realistically you could use fiber networking stuff with only one fiber


just leave TxD open...


System a emits telemetry. System b reads telemetry and exposes to the internet.

You can't drive my car simply by pointing a webcam on the steering wheel.


Sure, but think through the details. There are ways to expose telemetry through an otherwise air gapped system, but it is actually not easy to implement correctly.

My position is that if there's any connection, of any practical type, between the control system and the outside world, an attacker will be able to find ways to break things.

Even if it just boils down to "an attacker made our dashboards report numbers wrong and so we closed a valve to prevent an issue that didn't exist, and inadvertently destroyed our own product", the outcome is the same.


I don’t think the technical implementation is the barrier.

I would be much more worried about the human element of technology development. Namely that having a web based interface which shows data, but can’t change anything would feel unnatural. I can totaly imagine that you make the choice, implement it correctly. you have your website only showing things but unable to make changes.

A day later a product person asks for the ability to set the temperature on the web interface. You kindly and patiently explain to them that due to safety reasons you can’t implement that feature. They will understand and nod.

A month later one of the execs asks for a similar feature. They might also present the feature as mission critical, or even package it in the language of safety. After all customer X just lost so much money because we couldn’t do this one simple thing via the web interface. You argue with them and your viewpoint eventually wins. But it is a bit hard to argue because you ultimately have to argue that you / your team cannot be trusted to implement the feature safely. So you have to make it sound like you are both competent enough to be trusted with this hard decision and incompetent at making things secure enough the same time. A hard sell. But you win somehow.

Then a month later there is a hackathon. And one of the new hires puts together this amazingly looking new product UI. Everyone is amazed it is genuenly a delight to use. But it completely ignores the data diode principles. You have to kick it into the long grass, or modify it so much that it is barelly recognisable.

Then a month later a new product development start. Your company is hoping to conquer some new vertical, and need to make a cheaper version of the product. So an electrical engineer is tasked to cost optimise the control unit. They correctly point it out that we could reduce the BOM cost by more than half if only we got rid of the data diode and the second set of microcontroller. You have to argue the safety case. People nod. But you are loosing more and more “political capital”. Upper management asks you to work with the EE on how you can maintain the safety while also cutting cost. You know that this won’t be easy and the savings won’t be as much.

And so on and so on. So even if the company implements such data diode, and implements it correctly and safely. Future product and business pressures will work against it. Realistically you are not going to win every single argument about it. One day you will come back from a holiday and there will be someone who circumvented it either because they didn’t know better, or didn’t believe that they cannot be trusted to make the circumvention safe enough. It is a people problem, not a tech problem. If there is even such a distinction.


The salesperson comes in and says they've sold a big order of x units based on having a web portal that can set the temperature.


“hackers generated fake alerts” — harder than it sounds.


There're different devices. You put online some but not the others.

Temperature sensor is a sensor. Refrigeration compressor motor is an actuator.

GPS positioning of a car is a sensor. Motor to turn the wheel is an actuator.

Aircraft altimeter is a sensor. Motors to control control surfaces are actuators.

And so on.


Sure, but as stated in reply to a sibling comment, you then just hack the read-only system to mis-report sensor state and interfere with operations that way instead. Or use one of a million other approaches. Attacks won't follow your rules.


I think this comes down to the implementation of a particular system. But in general, since control system (the one that controls things through actuators) is NOT online, one can't interfere with it through online channels. They can interfere with sensors and fake readings reported online for example, probably, but not the ones used by the control system as it's offline.


Indeed, I agree that it's better to have the control systems separate from the telemetry systems and that absolutely does prevent certain types of attack. It's dangerous to think that prevents all types of attack, however.


The most straightforward way to make a fridge with monitoring over Ethernet is to slap on a Pi that reads off of the microcontroller managing the temperature onto an existing product. Most microcontrollers don't support raw memory access or reprogramming through regular interface, so faked values on the Web UI won't do much harm.

The most straightforward way to make a fridge with control over Ethernet is to slap on a 5-10 years old industrial temperature control product and wire up sensors and actuators onto it. In such cases, the industrial system that would be now 25 years by now would be running some cumulative 35 years old RTOS and Claude is going to find out it has bunch of basic buffer overflows that can make the fridge explode if exploited.


Difference between a smart thermostat for your ac and an Arduino temperature sensor next to your AC thermostat?


A firewall which allows to send sensor metrics from in internal network to an external collector and block everything else will go a long way. Not an air gap but much batter than whatever French supermarkets had in place.


"Simply"? You might as well say the same thing about creating a doomsday virus.


Yes, with AI I will be able to simply build a doomsday machine. Claude will just tell me how to build a supply chain, circumvent regulations and do this without anyone finding out.


Most of us effectively function as next-token predictors, in my experience. Some of us are trained not to dereference null pointers, but there's always a non-zero probability that we do so anyway. To reduce that risk I use the same tools an LLM would, i.e. code review, automated testing and verification, etc. There isn't obviously much difference between humans and LLMs in that sense.

Like you I would not vibe code something important, but I think that's primarily because I want to have a deep understanding of a system before I accept responsibility for it. An LLM would still be useful during implementation and likely to increase the quality of the result.

It seems very likely that AIs will become capable enough that they can take complete responsibility for development, deployment and maintenance of the systems we're used to building. Further, it seems likely they'll end up much more capable than humans have been. I think that's an eventuality we need to be prepared for.

A well-utilised AI tool can probably already build a much safer and more robust security/legal/user related system than you or I can, in any given time budget. There's a question around what happens as humans are required less for the "well utilised" part, and who is responsible when humans are not really involved anymore.


Guidelines | FAQ | Lists | API | Security | DMCA | Apply to YC | Contact

Search: