Schneier: It's Time to Regulate IoT to Improve Cyber-Security
eweek.com
eweek.com
There seem to be some soft mechanisms that governments could explore. Maybe something like demanding opening the source code once security updates for the device stop, so consumers could help themselves. Or at least, an even more modest regulation, simply allowing all consumers to hack their own devices without fear of violating any laws.
One thing that could hurt the market is maybe making manufacturers liable for the damages caused by security holes in their devices, but regulation doesn't have to go that far to make an impact.
That might help the readers of HN, but not users in general. Most users won't bother installing security updates for their PC if it's not forced on them. Updating one's light bulbs with something off github is a non-starter.
Edit: To complete the thought, it's not generally that hard to flash devices that support it, so a report that says "X,Y and Z have exploits, here are some options" could go a long way. Making devices support some minimum standard of local upgradability would help immeasurably.
Who's going to update the firmware in your light bulbs using a patched open source version of the manufacturer's code?
Edit:
> it's not generally that hard to flash devices that support it, so a report that says "X,Y and Z have exploits, here are some options" could go a long way. Making devices support some minimum standard of local upgradability would help immeasurably.
You're right, it's not hard. But can you imagine regular users doing that? Anti-viruses, Microsoft, Apple, Google and all major browsers had to automate and force updates on users to keep devices secure.
I have a hard time believing that any significant portion of users will flash updated firmware onto IoT devices to fix security vulnerabilities. Heck, when was the last time you checked for firmware updates for your home router? I do this stuff for a living and I don't think I've updated my (current) router's firmware more than once in the last few years.
If you've already got a database of router versions, it's not hard to include how to submit a new firmware version in that database, whether it be a POST URL, any params and the expected payload(s) or a TFTP upload. At that point, the same device that generated the report could give you a management page that listed some options and semi-automated the process. Want to update fridge with out of date/exploitable firmware with community fridge firmware X? Click here. Want to update with community firmware Y? Click here. Want to update to newer/latest proprietary firmware? Click here.
I do agree it's not a solution to the problem. But it might help. It would also allow your techie friend to run an app on their phone or laptop when they come over and let you know, or handle it for you with a minimum of fuss.
This problem won't ever get better until someone starts being held accountable for exploitable network attached devices. I think part of the reason that doesn't happen is because it's not feasible with current norms of behavior to expect producers nor consumers to do so with any level of confidence. Providing tools for this may take us one step closer.
Personally I think a blend of regulations will be required. If devices sold are required to be updatable in one of a few ways, and workable firmware open sourced if obsoleted (or clearly marked or all labeling as may be entirely unsupported after 20XX), then it becomes feasible to require users to have some level of accountability over what they put on their local networks. I think it's the same as cars and the public road system. Cars have to adhere to certain standards to be street legal, and car makers must adhere to certain standard to sell their cars as street legal.
> Heck, when was the last time you checked for firmware updates for your home router?
Every 6-12 months, generally when I'm checking why performance is bad at the moment. But I'm not getting a report, so I only check when it's on my mind.
Ideally we would have this suggestion in addition to a stick to ensure people have an incentive to sell relatively secure hardware.
We have them for practically everything. End users can't figure it out, the manufacturer (say Microsoft) doesn't give two craps, so a third party (>90% of my career) involves solving those problems for users.
If things were open sourced, that'd make my job insanely easier in terms of man hours invested in any project.
I've literally had to spend >100 hours diagnosing that Dynamics CRM 2013's main data import tool has a bug that they refuse to fix. (Ever since it launched.) The import wizard refuses to actually load TIMESTAMPS (as well as some other columns) in even though it specifically supports setting that column. So when you do a data import, ALL of the user's data is out of order. That's completely unacceptable for any business to have their e-mails, notes, all have the same timestamp and no proper ordering. So "all" you have to do is first diagnose that this is the problem, then find this super obscure website where a guy made a fix except his code is half completed and full of errors, then install the entire CRM SDK stack. Learn how to write a C# plugin that exploits the CRM SDK. Write the corrected version of this guys code. Compile it. Use the CRM SDK toolkit to "attach" the plugin to your CRM. Figure out (no docs) what messages/events the plugin is supposed to actually trigger on with no error messages (while it takes >5-10 mins of user input every test to find out if it worked and then remove all the data again from the DB.)
If that was open source, that'd be a one line of code fix and I wouldn't have to play the "prod the black box" game of trying to figure out exactly where the failure point is. Hell, just adding a debugger would be insanely powerful. And that was just "one" story I have with "one" product that I support out of dozens if not hundreds.
So that's the thing. Regardless of whether a manufacturer open sources or not, people like me still have to fix the damn problem. Failure isn't an option. The checks have to go out. The lung machine has to keep blowing. And whether or not the business makes it easy or hard, doesn't enter the equation (except in terms of cost and man-hours). Nobody cares that you want to protect your IP. They care that their damn time clock doesn't work and they've got millions of dollars of product to ship while their employees stand outside of a locked automatic door.
So the real question is simply: Do you love, or hate, the people who support your software? Because we're still here, every day, slogging through insane problems by poking blindly in the dark. Closed source software breaks just as often as open source. The difference is, that the essential third-party (me) has to bill 10X as many hours to fix the problem. So if you're a business, it's in your best financial interest to demand open-source code whenever not prohibited (security constraints, HIPPA, whatever).
Closed-source is like buying a car and pretending it'll never break down, or that the manufacturer has the time, energy, and willpower to come down to your house and fix it. Some might go that far. The other 99% won't. And if there's a piece of software that has zero bugs, I haven't seen it in my lifetime.
So to bring this back to IoT. When a clients IP Camera goes down, they don't give a crap why. They need it back up. Yesterday. And they especially don't appreciate buying expensive hardware that ends up having security flaws... and worse... that is physically impossible to close. The clients aren't going to easily understand, "I paid $3,000 for it, and it works fine, but I should NEVER use it because it could be hacked?" That's like buying a car and a recall hits that says "the locks no longer work" and instead of fixing the locks, or allowing a third-party to fix the locks, everyone is just supposed to stop driving their car. You'd never see that in the real world, but somehow "software is magic" so it plays by different rules.
Try walking into Home Depot and asking the salesperson how to update the firmware on the lightbulbs he is selling. You will get your sanity questioned, especially once you explain _why_ it is important to update.
What reasonable case is there for making them not liable for the damages caused?
Surely the right idea in principle, though. I'm just not sure how realistic is it to implement in a smart manner.
After what period of time? Microsoft updated Windows XP after the official end-of-life of the OS for consumers with the patch to the Samba protocol to prevent WannaCry, but if they hadn't bothered, would they still be liable? Should they?
Are they obligated to update pirated versions of their OS?
Is the manufacturer liable if they release a software patch but the product owner doesn't apply the upgrade? What share of the liability should each party take?
Software is largely immune to liability litigation in the USA because the current legal status of it is not legally "a product". Converting it to "a product" for the purposes of liability is a major sea change for our understanding of what business models can be applied to software, licensing, ownership, etc.
Also, does the average product programmer carry some sort of programming insurance? Are we going to force every web developer and every open source programmer to carry insurance, to be licensed to program, and to live up to specific ethical standards?
I'm not saying these are undesirable changes, just that they are changes and there are a ton of issues programmers don't foresee that should be discussed before making
They probably don't even have the (usable) source code.
https://www.arm.com/news/2017/10/a-common-industry-framework
That's what insurance is for. Something like this was discussed in my torts class in law school, except that was long before IoT devices existed so it was about things like lawn mowers.
The idea is that it might make the most sense economically to make the lawn mower manufacturer liable when users cut off a finger or toe, even if it was due to consumer stupidity instead of any negligence on the part of the manufacturer, because the manufacturer is in the best position to estimate the risks and purchase insurance to cover them.
Presumably, the manufacturer will pass on the costs of those insurance premiums to the consumers. The manufacturer still has an incentive to try to build safe mowers, because if their insurance company ends up paying out a lot the premiums will go up. They can pass those higher premiums on to the consumers, of course, but that will make them more expensive than their safer competitors.
The manufacturer is in a good position to deal with insuring for these injuries because they know how many mowers they are selling. Other candidates have less useful information. For example, a consumer usually has no idea what their chances are of suffering a lawn mower accident, so have no idea how to decide how much insurances is needed. Health insurance companies will have a good idea in the aggregate of how many of these accidents occur in a year, but they have no idea which of their customers use lawn mowers.
Done right, this should not hurt a market much.
I'm not sure it could work for IoT, though, because a lot of IoT devices are made by new companies that probably will not be around long. With things like lawn mowers, you could takes years to get around to cutting your hand off, and still reasonably expect the manufacturer to be around. Not so with a lot of IoT devices.
Freedom Markets™ are best served by binding arbitration.
My comment about this got 0 attention though...
What are the outcomes for using insurance, and what are the outcomes for using regulation? Does anyone know the answers in a technical policy sense (not in a philosophical sense)? They are different tools useful for different problems.
Thinking out loud, insurance seems like a poor solution when people will suffer serious, irreparable harm. If the lawnmower severs a foot, then an insurance payout isn't really sufficient; regulations should prevent that from happening in the first place. More broadly, my point is that there are larger issues than economics.
> Presumably, the manufacturer will pass on the costs of those insurance premiums to the consumers.
In economics, that is not the case. Businesses don't price goods at 'cost-plus'; they don't look at their costs and add a profit margin. Think of the soda at the movie theater on one hand, and on the other the car being sold at a loss because the market is soft; think of software. Like all businesses, the lawnmower manufacturer already is pricing their product at the level that maximizes total sales revenue, which depends on supply and demand; raising the price will reduce revenue (probably because unit volume will decrease too much to be compensated for by per-unit revenue increase). If they could raise the price and bring in more revenue, they would have done that already.
There is an issue of elasticity: If your customers' alternatives are limited - i.e., if they can't go without your product, if there are limited competitors and substitutes - then you can raise prices more easily. A Van Gogh painting is highly inelastic; lawnmowers are much less so.
Bad question, false dichotomy: these are not apples-to-apples comparisons. How about some minimal regulations that include an insurance requirement?
So I agree: yes, the existence of tort law sometimes obviates the need for regulation; but has hardly banished regulation or the need for it entire!
Eh, more or less. Increases in the price of supplies sometimes do lead to an increase in the end price of consumer goods. I agree that this is more likely to happen for inelastic goods, though.
What isn't being mentioned here is that insurers will require insured companies to do a bunch of stuff in order to remain insured. Have some processes in place, do some things, and so on.
Just like your car insuruance isn't valid if you drink and drive, your software company insurance might not be valid if you aren't using source control and have no testing or code review proccess.
So in some ways, what you get out of insurance is market driven 'regulation'.
Possibly, because a rival insurance company can impose different conditions, insurers are incentivised to require only the stuff that really reduces risk, while regulators might make irrational regulations driven by moral panics in the press. Similarly, they might be more responsive to supporting new technqiues that reduce risk at lower costs, because customers will seek out an insurer that lets them use those.
Possibly, insurers are less suceptible to moral hazard, where regulator employees have close relationships with industry heavyweights and make policy that helps their buisness, at the expense of consumers and sector rivals. Insurers mostly want to make money.
Regulators might be better sometimes because their staff tend to be very mission driven (they want to fix the problem, not make money). They will get a kicking from the public if there is a big accident and laws and regulations did not prevent it, while an insurer is only punished if they don't satisfy current law. So a regulator might be more proactive.
I'd only add that the insurer's incentive to reduce risks to their profits is not always aligned with consumer's risks. As a simple example, the insurer may require the vendor to have consumers sign away their legal rights to reduce the risk of expensive lawsuits.
If we regulate insurance so that a product needs to be backed by a 10-year liability insurance, it might have 2 good outcomes to fix this:
1) the insurance company might ask a lot of hard questions and require audits, plans, etc 2) the insurance company might require escrow to the iot device's update mechanism in case of bankruptcy, so they can remote-brick the devices
It's unlikely that the reality would be so rosy though - mainly the claims against the insurance would likely be too small in many cases. It might help in 10 years against the biggest ddos botnets. Not nearly as good as direct regulation.
It sounds great if I'm MegaCorp with an entire department for dealing with this stuff, but as an independent developer trying to start something it's a horrible idea.
This is the problem.
And "solutions" like that can easily backfire, because they squeeze medium sized companies out of the market, who might stake out a middle ground between $20,000 Cisco hardware and $150 "disposable cameras" from Fly By Night Corp, by charging $200 for a product that actually guarantees patches for a specific number of years.
Because the liability risk raises the price the middle-ground company has to charge to the point that they can't compete with Fly By Night Corp who doesn't care. Which means more sales of insecure garbage products, because the customer no longer has the option to buy an affordable product that receives patches, and that customer can't afford the $20,000 industrial grade product, so more people end up with the insecure garbage.
It would be much more effective to just require manufacturers to provide security patches for e.g. 7 years. Fly By Night Corp still wouldn't do it, but then at least Samsung would.
Then they will sell two versions of the same product. One lets you voluntarily waive that requirement for the same price as today, or a "normal" version where it costs orders of magnitude more.
Or at least, an even more modest regulation, simply allowing all consumers to hack their own devices without fear of violating any laws.
More simply, they could make all reverse-engineering legal, but then the copyright/IP lobby would strongly oppose.
Unless said regulation was the ability to actually own your device and install your own firmware, doubly so once it becomes unsupported?
Put pressure on ISPs to make them take responsibility for traffic originating from their networks. They have the the tools to notify customers if customer is sending suspicious traffic (and if necessary, they can temporarily shutdown the connection).
Copyright law is not in agreement with this. Reasonably, a manufacturer could argue that follow up installments of the software are still actively sold. Then one question is why the fixes wouldn't be ported back to the old release and I guess because differences in the code require significant effort to adapt the fix.
Regulation should seek to equalize ...
> manufacturers liable for the damages caused by security holes in their devices
This is not easy if a chain of bugs in different programs is used to create an exploit. And it would be damaging to warranty wavers especially in open source. After all, the server side will likely run full fledged open source stack. This is where service providers step in. Google would probably like to do monitoring and instrumentation as a service, among others.
It's interesting to see how the US citizen ordering the product of a back alley company in Shenzhen via Alibaba is supposed to recoup his damages. It's impossible, period.
With wired IoT devices: segment your home network, always use a trusted gateway application and never allow your IoT devices direct WAN access. With wireless devices, all bets are off, since you don't know at all who can access them (i.e. they can interpret WiFi frames in unassociated state directly in silicon, and you'll never now, even if your stack is completely open source) if they are in the vicinity.
For one, it's time to hold companies (and executives!) accountable for security of the data they are charged with protecting, often without your consent (eg Equifax).
For another, insufficient product liability for companies being lax--even negligent--with security. Honestly I don't see an outcome like network-connected lightbulbs bringing down the Internet as particularly far-fetched.
Frankly I don't even know what the market for IoT even is. Who needs $50 light bulbs that will DDoS someone one day? Or, worse, compromise your network to an attacker.
And all for what? So you can turn the lights on after you go through multiple steps to unlock your phone?
I wanted network-connected lightbulbs so I could have them turn on at the time I needed to wake up, when that time was well before dawn.
I never installed them because I didn't know how to secure them and my schedule got more reasonable, but I think the use case is pretty compelling.
I mean, an Internet connected light bulb use-case would a bulb that flashed whenever a stock you owned went down in price, which is ridiculous despite being the least ridiculous example I could think of.
IoT security cameras and an automated kitchen you phone to, to prepared you dinner if you were coming home unexpected seems like the least crazy IoT device an individual could own - most IoT stuff seems more like what a global company would want rather than an individual.
A lighbulb flashes when visitors ring the doorbell, which is useful for people with visual impairment. the doorbell has a hidden rfid reader, and certain guests have an rfid card. the doorbell flashes differently for each visitor.
Take it one step further, and you could trigger it via an NFC read of your vistors' phones. This also would not need to connect to a network (beyond installing the app on the phones).
Actually, sounds like it might be a fun project to play with :D
How is this an example of something that needs to be connected to the Internet. My point is most examples are about local connect to justify a jump into a world-wide, insecure net.
I mean, all the light bulb examples are contrived because a PHONE has all information-conveying ability of a light bulb and is designed for information-conveyance. Light bulb are designed for the light-consumption needs of those nearby and examples of light-bulb-control at a distance are either poor substitutes for phones or "weirdness" - haunted houses and art-happenings.
that's well and good until they cause negative externalities
I already have a network connected to the internet. I don't already have a dedicated light bulb timer, and producing one of those, or buying one that had somehow gotten produced, would be stupid.
I had no particular desire for the bulbs to be connected to the internet, but I did want them connected to my home network, and I did also want my home network connected to the internet.
You get a combination of a wifi network being the one sort network every person is guaranteed to be already on and every company dreaming of cloud-based-app supremacy "synergizing" every useless whatsit they have access and the IoT nightmare is in full swing.
"What could possibly go" - wait, we've seen answers to that one already.
The fact that they're IoT devices is kind of irrelevant, as is their function. The underlying question is: to what extent should manufacturers be made responsible for the damage their products cause?
If you make an electric heater that routinely combusts during normal operation, your customers and the state have some recourse against you – not that it would happen, because there are generally standards in that industry. But if you make a device that quietly becomes part of a botnet, you really aren't going to suffer; even the reputational issues are generally minimal.
I don't get it either. I can get the idea of switching on and off all the lighting in a home e.g. when leaving (hotel room style) but to do that you need to have a hell of a lot of gadgets, not just a couple of "smart lightbulbs".
The only reason we have smart lightbulbs for switching to begin with was because someone realized that the sockets are standardized and users can switch bulbs.
The elephant in the room is the wall switch, it's now in series with the bulb! Every time someone switches off the light using the lightswitch, the smart bulb can't be switched on again! And no - people will NOT stop switching lights manually just because they have a smart bulb!
What people really want, would be smart switches, not smart bulbs. With a smart switch, you can switch either by hand or remotely.
I just don't get why manufacturers (and customers, most of all) don't realize this. Are people happier to buy the bulb because it's a 10 second job to change, compared to the wall switch which might be an hour worst case require an electrician?
I'd be thrilled to change all my switches and dimmners to smart ones, once there are gadgets that are cheap, secure, reliable etc. I'd mostly use it to have a master lightswitch somewhere near the front door.
There are several problems with regulation: a) Whack-a-mole, new ideas and business models arise faster than the speed of government. b) Regulatory capture, like what happened to our banking regulations. c) More often than not, penalties are captured by the regulator, but compensation is not made to the injured parties. d) International law / trade agreement complications. I'm sure there are many more.
If manufacturers knew they'd be on the hook for damages in a dollar-to-dollar way, they'd put more engineering into their work, they would price it accordingly, and personally I'd be fine with that.
Regulation is not blind, but it does raise the floor.
I would figure large service providers, and big companies with fleets of marginally protected devices, would be the ones to bear most of the damages coming from shoddy IoT devices being pwned and rolled into larger ddos/ransomware/etc attacks.
I just think broadening the concept of liability in our industry to the point where some company's negligence is finally made an example of has a much better chance of improving things than a panel of regulatory agency appointees that have either already been paid as industry lobbyists, or are operating under the carrot of a future position as a highly-paid exec--just like the FCC, or Robert Rubin, or any of Trump's appointees, or so many others...
In practice I expect it to result in fewer products on the market that are more expensive and no more secure as this sort of regulation will simply select for large companies who are experts at paperwork and soft bribes.
I wish I had a better idea.
Similar to costco - sometimes they do not sell products that they believe don't have the quality their customers deserve.
What do you mean "this sort?" Schneier just says "we need regulation", that's pretty definitely all the article says.
Something that already works are various forms of certification.
Examples are:
* "Norton protected" on websites
* Underwriters Laboratories on US products
* US DOD Trusted Computer System Evaluation Critera for how the US military checks the security of a product
* ISO 9001 for quality management
* Oregon Tilth for certifying organic products
Some of these are more valuable than others, but are at least a separate stamp of approval.
With my magic wand I'd have a group come together to form such a certifying organisation and provide it with a marketing budget. The marketing would mirror the success of "check for a green padlock on a website to know it is secure" - look for the stamp on a product box.
If that doesn't happen and there are more serious incidents is that consumers would look to Google, Apple, Amazon etc. eg if the proposed device doesn't work with the Apple hub and been vetted by Apple, then they wouldn't buy it.
So yes, regulations are circumvented or ignored intentionally, but that doesn't mean that it's somehow a bad idea; the idea is to enforce positive behavior by attaching an undesirable response to not doing the positive behavior. It's meant to shape behavior in the long run, and it's why pulling funding and reach from regulatory agencies is usually shitty, as more than anything, it's meant to induce a scenario where one can say "look, we have these regulations, but X, Y, and Z continue to do the bad behavior! The regulation is senseless!"; in reality, X Y and Z have just run the costs and determined that it's cheaper to fight the regulation than to adhere to it.
It also isn't necessary for every consumer to care about certification. All it takes is a minority to do so - enough to tip sales away from the competition. Certification is also easy to mention in reviews and product feature lists.
Without effective regulation I imagine you'd see aircraft falling out of the sky left and right because - and let's be honest here - safety is probably at the bottom of the priorities, both for manufacturers and airline operators.
Most people don't believe that accidents can happen to them. People rationalize the risk away, so they skimp on security to save a buck.
So, operators don't have incentive to maintain and upgrade their aircraft because that has only a negligible effect on performance, but costs a ton and the perceived risk is low. Manufacturers, likewise, have no incentive to produce safer airplanes because safety doesn't sell: range, capacity and fuel efficiency do.
Yet, air travel as never been safer. Seems to me like regulation works ok in this case.
The bottom, really? That's a pretty bold statement to make, and a disservice to the work those have done to ensure we have thousands of safe flights every single day. I'm certainly not arguing against airline regulation, but I don't believe that's the only thing holding them back from crashing planes on the regular.
Neither argument seems particularly persuasive without being falsifiable.
Which are exactly the kind of companies who can do good engineering, so why wouldn't they do that instead?
The only way to avoid a botnet apocalypse is to secure home routers. Outbound traffic should not be an automatic right. People should have to authorise each device for each type of traffic.
I argued this long-hand when OVH was taken down by "security" cameras. Good to see we've made no progress.
https://thepcspy.com/read/when-did-we-stop-caring-network-se...
Nobody wants to do this. I'm a quarter way paranoid about electronic security and even I don't want to do this.
I recommend a certification programme, with manufacturers justifying the access their device needs. A little signed JSON blob of hosts and ports it plans on connecting to. The device communicates this to the router. If the signing certificate is still valid and the manufacturer trusted, the router could just allow that access, or prompt the user to just let them know that device is trying to connect. No confusing detail. And just once, at the same time you're setting up network stuff, so it's not weird or extra hassle.
It's leaps and bounds better than what we currently have. The vast majority of us have zero idea what the devices on our network are actually doing, all while we're each throwing dozens of these cheap internet-enabled things online.
For legacy devices, a more iterative approach might be needed but it can still be prompted: "Dell computer is trying to connect to clearlybaddomain-dot-com. Allow, Allow All, Deny, Quarantine". You could even layer on some "known bad" hosts or traffic patterns via centralised lists to automatically quarantine devices at the router level.
Nothing here is rocket surgery. One developer for a few months. An entity like Google could do this in an afternoon. There's just surprisingly little appetite for it.
Seems somewhat unreasonable to expect people to know what traffic a device needs, and if you ask too much, people will just default to allowing it without even thinking about it (I still occasionally see people suggesting using a DMZ to make online games work, rather than forwarding the necessary ports [and even that seems rarely required]).
A few examples: - It's currently pretty hard to add HTTPS on a router admin page. - Browsers can't do service detection on a local network, so we have to resort to central servers to manage headless devices (or ugly, unreliable local IPs). - Punching through firewalls / NAT is still hard, so we again resort to central servers.
It's really fucking hard to do IoT at scale, in a easy-to-use way that's secure and that respects the user's privacy. I think we can solve that.
I guess it only took this long for enough people to become insane enough to justify a market for it.
Preach on brother Bruce!
If I want to buy cheap hardware or software that isn't certified I should be able to.
You have a right to pose a danger to yourself. You don't have a right to pose a danger to others.
In fact it is then inconsequential whether some device puts you in direct danger or in danger of somebody comming after you for putting them in danger.
Right. So you, the little guy, is going to shift the blame to a rich company that does this for a living. Say it's a company such as Google which makes some IoT devices. Would you be able to prove in court, and in front of their engineers and lawyers that they've sold you an insecure device? Do you even have access to their source code?
Maybe our current situation is what it might've been like when we had devices that could generate and receive RF emissions but before there was an FCC/global regulatory distribution of spectrum.
If that analogy holds at all, perhaps that might be why it should be illegal to manufacture or import devices which don't conform.
Both such devices do as much if not more harm to your neighbor as to you. An electrical appliance that interferes with TV broadcasts may not bother you but it bothers your neighbor. An IoT camera that's hacked to DoS a hospital may not bother you but it bothers the patients of the hospital.
In some fictional Libertarian wonderland, these problems could be dealt with neighbors suing neighbors but in the real world we need the state to regulate products with serious cost "externalities".
As soon as they are successful, they'd be breaking the law.
Just make the owners of IoT devices liable for damage they cause.
Frankly, I'm tired of trying to find time to find specs (and then learn how it really works) to figure that out for myself; I secretly dread receiving anything networked for the holidays. It's difficult for me; most end-users have no hope of protecting themselves - and they seem to assume that any product sold must be safe. They assume it is regulated, in effect.
[0] Arguably the Internet is physically different than spectrum. Physics 'creates' the spectrum and its physical limitations make it a public good; there are only so many frequencies, and propagation is part of the equation. The Internet is a creation of humans and in theory can be recreated or modified at will. But that theory isn't realistic: The Internet cannot be replaced or substantially modified; the public has no realistic option of using a different Internet if they don't like this one. In any practical sense, it's a public good.
Before the iPhone, software and hardware were often different and had different business models.
Give it 5-10 years when enough of them have died off; then change will happen.
That said, I've more or less completely ignored IoT so far aside from passing interest in how easy Mirai was and I've only briefly dabbled in firewall configuration, so many of my assumptions could be wrong.
Separately, no - attacks like CSRF will quite happily be routed and compromise an incompetently designed IoT device.
CSRF has been around since 2001 and is in the OWASP top 10. It would be absolutely valid for regulators to require reasonable steps to be taken to prevent its abuse, along with similar attacks.
This of course doesn't do anything to secure attacks via one of those trusted addresses, but does prevent someone just happening across an open device.
There's also the problem of coping with dynamic addresses, but that can probably be handled separately.
Routers do have a solution: some support setting up a VPN, to which the phone could securely authenticate against. But good luck getting users to configure that.
Or they can be attacked over the LAN before the router gets a look at the traffic.
I mean, I program embedded systems for a living - and I couldn't tell you which IoT dash camera or digital camera with wifi or internet-connected car entertainment system is secure.
Instead, I feel that accountability would work much better here. If you're selling an IoT device, and you haven't taken industry standard precautions for securing it, then you're on the hook for whatever your device is used for. The same can be applies to companies storing personal information e.g. Equifax.
What happens when you want to sue a company for lack of updates when the company went out of business 6mo after it was created (like Juciero)? You can’t sue them, where a law could have forced them to be secure from the start or put up a bond to support the devices for a while.
Companies will do the bare minimum to get certified? You realize that’s a massive jump compared to what happens now.
That's still quite a chilling effect though. Well, maybe it should be. Now we're really careful about what we buy. But maybe it's too much. Who wants to expose themselves to a small chance of high liability? Okay, so allow insurance against said liability. Wouldn't that defeat the purpose? Well, no. Insurance companies would only insure against liability for devices they've vetted and approved.
Boom, market regulation. It works for cars (see IIHS).
I don’t think I’m qualified to try to do that for something I might buy, let alone ‘normal’ people.
That’s one of the problems with the market approach. The information assymetry is so big that it’s not a reasonable demand on a person for a $15-20 lightbulb.
I can't brainstorm many simple solutions that blanket cover many different types of businesses, markets and scenarios.
However, give the certification process some thought too. I can see quite a few drawbacks here as well.
First, a significant advantage for established, rich companies. We'll be swamped with IoT from Apple, Google, Facebook and Amazon while small competitors have a hard time getting their products to the market.
Second, you'd need give the regulatory body access to both your software and your hardware. And what if the device is connected to some cloud server? That body may need to look at its code too to make sure that your control server is compliant. And what about the network? The database? Where does it end? And do you need to re-certify each and every version of your server? What if you introduce a security vulnerability?
Third, certification can't be a one-time deal. That protocol your lightbulb uses to talk to the microwave oven for whatever reason? Well, someone broke that and can now make both of them divulge your dirtiest secrets. The same regulatory body would have to keep track of such vulnerabilities and force manufacturers to update their devices - and what if the manufacturer has gone bankrupt? What if he doesn't want to update these devices? Are you going to force people to throw away their light bulbs? You'll have to, otherwise you're back to square one in which all devices are compromised, only now it takes a little bit longer.
Imagine the bureaucracy all of this will require.
I imagine a market would appear for some of the basic software (Linux diaries, etc) to help make things easy for small companies that do want to do it all themselves.
I like the government idea because frankly I can’t think of anything else that would work (outside a rediculously improbable change in consumer behavior).
That's what I meant by check-mark security. Yes, it is better than nothing, and by all means let's do that. It's low hanging fruit, and it should be plucked. But in the end it amounts to little more than hanging an air re-freshener on a huge pile of garbage.
I'm just pointing out that such certification might cause executives in companies that today put more effort into securing their devices to stop putting in that effort. If it's all the same to the consumer, why spend any more than the bare minimum to get a check mark?
Today there's no clear bar, and a good engineering team will always be able to convince a responsible management that they need to put effort in security. But once that fairly low bar is set, I think that the next order from the management will be "make our devices certifiable and nothing more".
Checklist car safety means a pretty safe car these days, and the companies that go beyond do amazing things.
I'm surprised PCI (payment card industry) security standards have not been mentioned on this thread. There's a case where non-government regulation has, by some definition, worked.
PCI isn't perfect: if you check all of the boxes, it doesn't mean you're secure, but I'd rather have the industry self-regulate than have politics come into play.
IoT router/firewall might be one of the solution here, i.e. adding IoT pattern into existing routes/firewalls to protect IoT devices, in addition to your PCs and sometimes BYODs(smart phones etc).
It is very hard to make all IoT devices secure due to limited resource they have, so the first line of protection should be done on the router/firewall/gateway I think.
I of course hope all firmware will be safe, and they should be safe as much as possible, still, you need a more powerful device to safeguard them. Put another way, no matter how secure my wifi-bulb is designed, I'm not going to expose it to the internet, and I will put it behind my firewall/NAT-router.
Why?
> you need a more powerful device to safeguard them
Why?
> I'm not going to expose it to the internet, and I will put it behind my firewall/NAT-router.
Why?
Sure, the idea of billions of devices around the world connected somehow is scary, but government regulation is not the answer. If anything, regulation needs to be decentralized. More open source, community involvement with reviews and discussions, more self regulation
Regulating IoT is tougher, but is analogous to licensing the airwaves.
For example, someone could create a company that issued certificates of security. Manufacturers would pay a small fee to these companies to perform security tests and give them a certificate of security. They can put that label on their products to provide confidence to consumers. Some products may warrant a much higher level of scrutiny than others so there could be different levels or different companies that offer it.
I think people will naturally choose the products that are 'certified' over the ones that aren't, and manufacturers will have to end up doing it to stay competitive.
USB Type-C can deliver enough power to seriously damage your $1000 MacBook if the cable/adapter is designed poorly. There is a certification process, but most products on the market are still below-par. Below I will link a list. Guess what, those "bad" products are still bought en masse.
This week, I discovered that pretty much all the water filters that are popular for the type I'm looking for aren't even certified to filter out harmful materials. NSF 53 certification exists, but it looks like the market didn't do any research into it and trusted NSF 42, which was touted but is a much less strict standard, filtering out odor and taste (important in its own way). Theoretically, these filters could be passing on lead and asbestos.
Your solution _might_ work for a _part_ of the market, but it is almost guaranteed that there will still exist a significant (if not majority) part of the market, that doesn't care about certification/prefers the cheaper product.
https://docs.google.com/spreadsheets/d/1vnpEXfo2HCGADdd9G2x9...
What makes 'IoT' any different from an ordinary network-connected computer? You're either saying "it's time to regulate networked computing devices" or, "I want to carve out an easygoing regulation-free niche for MY product[s] to artificially excel in."
I try not to be needleslly pessimistic, but this article has no definition of 'IoT' beyond 'networked computer with sensor', so three guesses as to which one it is.
How do I update my lightbulb? Who will make updates for? Maybe Phillips will for their product but what about smaller OEMs? What if the company quickly goes out of business like Juicero?
Depending on what you buy and where you buy it do people even know who made it? Would you even know how to check for updates (assuming they exist)?
It’s also FAR more secure. IoT devices are often easy to hack. And while they may not have much horsepower they have a network connection. You won’t mine many Bitcoins but it doesn’t take a lot power to be part of a DDoS.
And I have one computer, one tablet, one phone. I may have 5 smart lightbulbs, a DVR, a security camera or two, a indoor/outdoor thermometer....
The phone is probably a bigger concern at scale, but I have seen plenty of families with dusty "photo storage/backup" boxes that their family's resident IT person set up and networked when they were in high school.
Basically the same things that make, say, 'rats' different from an ordinary human being: specialization, capabilities, defenses, and deployment density.
A different response might be needed to deal with millions of rats moving into your town compared to what worked when 15,000 juggalos congregated for the ICP music festival.
They're both 'mammals', true; they both sleep and eat cheese. But they still might necessitate different strategies to manage or cope with them.
(Having said that, I am still deeply skeptical of governmental regulation as a solution. I think Schneier is right about the scope of the problem, and that it's a perfect example of the class of problem that markets can't fix. Although trying to hold device makers liable might help somewhat, I'm afraid that the problem is like global warming — in theory, there are various solutions that might devise, but in practice human societies, even the minority of them that might have their shit sufficiently together to address the problem on their own, still aren't capable of the level of coordination required, so what we really need to do is "get ready to deal with it, because there's no way to stop it".)
Forget about hacking your car, what about the hacker that hacks a car fleet? What could a hacker do with a botnet of cars, each with cameras and maybe even face recognition. How about killing off people for the highest bidder at the push of a button.
So what about the old iPhone 5 and MacBook Pro from 2008 we gave to our kids for music, YouTube and getting started with computers?
The professor‘s response “this is just like all of those internet hit pieces”. Hmmmmm.
It's sort of like living in a neighborhood and having a rock pile you enjoy the aesthetics of, but know it's prone to having rattlesnakes move in, and instead of fixing the rattlesnake problem either as it happens or at the root, just putting a wall around your property excluding the pile. Sure, you're mostly safe, but when animals/children get bit, your solution starts to look quite a bit worse.
It will be more complicated than just 2 networks and it should be based on behavior and trust rather than device type. Consumers don't yet have the tools to monitor/categorize devices based on behavior, but some corporations (like mine) offer platforms that automatically move risky devices to network segments to mitigate risk.
Far from bulletproof though - it limits risk, but does nothing to prevent those trusted devices from becoming infecting and having essentially free-reign (which when you use Windows as your main OS is far from a theoretical concern).
[0] Introduces a double-NAT, which isn't ideal, but hasn't caused me any issues in practice.
Honestly, I don't understand why consumers lack the restraint to simply not buy unfinished products, but this is where the leverage to improve IoT security has to come from. If today's IoT devices are such a liability, then prove it in court; but don't think that you can write a law that ensures security instead of mere standardization.
Meticulously studying the introduction and effects of regulations in this vein, I highly doubt they will have the intended effect.
If we are going to regulate, we need to improve laws for consumer electronics across the board, with all the big players on board and participating.
More realistically, legislators should be crafting incentives. Establish liability statutes. Carve out liability exceptions for companies that can show they used industry best practices, hired engineers that are members of professional/industry organizations, pay for ethics training, establish good faith effort of security development (unit tests, integration tests, traffic encryption, encryption at rest, well designed key exchange architecture, software/firmware update architecture for at least X years after sale, etc).
Conversely, if a company does none of the above, it's easier for a consumer or an Attorney General to bring a case against the company, even if it's years too late to be useful.
The problem with litigation (as well as "free-market" solutions) is that this generally doesn't happen fast enough. The security damage of IoT will be externalized extremely quickly. Being able to identify and sue a foreign company that is far up the supply chain from the end-product.
First, yes of course automobiles are regulated and should be. The fact that some things that are and should be regulated include embedded internet hosts does not mean that all devices that include embedded internet hosts should be regulated. This is basic logic.
Second, holding one set of botnet victims responsible for the harms suffered by another (overlapping) set of botnet victims is perverse. Every host should be "secure"; very few are. A secure host wouldn't be a victim of a botnet, either by donating processor cycles or receiving unwanted traffic. If lawyers really need a job security program, consider that ISPs have lots of money and they actually could reduce DOS attacks; why not hold them responsible?
You could hold ISPs liable, and they would block untrusted IoT garbage at the network level. Or you could hold IoT garbage producers liable, and they would make security changes and/or pay ISPs to do some firewalling. Either way, the costs and results will probably turn out about the same.
The best solution, of course, would be having fewer things uselessly connected to the internet.
How on earth is a civilized society going to keep "useless" devices off the internet? Who gets to decide what is useless? For example, medical devices are notoriously insecure: what politician is actually going to get behind an effort to make Grandma's life more inconvenient and also shorter just to satisfy some nerds' idea of a perfect internet?
The nerds on HN disappoint. When faced with a hard problem, instead of doing the hard work to fix it, they want to involve the lawyers...