11 ms·
The sad thing for me, and I see this way too often, is that someone, somewhere, no doubt said "Of course its secure, you can only READ the CANbus, the software
by ChuckMcM 11y ago
The sad thing for me, and I see this way too often, is that someone, somewhere, no doubt said "Of course its secure, you can only READ the CANbus, the software doesn't even HAVE a write capability!" and everyone in the room nodded and went on with the rest of the review.
Manufacturers and engineers have to get it through their head that IF you can change the firmware ANYONE can change the firmware. If the firmware is SECURITY CRITICAL then the only way to change it can be through physical presence, loading encrypted and and signed firmware, with external validation. (something like the car asking for a third party to authenticate the operation ala nuclear launch codes). You can still get screwed but it will be hard enough to do that otherwise low value targets will remain relatively safe.
- csours 11y agoAnd when manufacturers start doing that, people will complain that they don't "own" their cars anymore. [1] 1: http://boingboing.net/2015/05/21/gm-says-you-dont-own-your-ca.html http://boingboing.net/2015/05/21/gm-says-you-dont-own-your-c... Disclaimer: I work for GM.
- saulrh 11y agoKeep a suitable private key for signing firmware in the car, locked in a secure box that can only be unlocked with two copies of the car's ignition keys? I'd call that sufficient security, but it'd also clearly allow me to get to the keys if I want.
- matteotom 11y agoThere is probably a middle ground here. For example, lock it down like the parent comment suggested, but have a process to unlock it, requiring (for example) physical access and some vehicle specific file from the manufacturer requiring proof of ownership and a scary warning. (I believe some phone manufacturers had a similar process for unlocking bootloaders a while back.) This would still allow enthusiasts and security researchers to "own" their vehicle and would keep malicious actors out (for the most part).
- pki 11y agoWouldn't this basically make cars no longer street legal?
- krasin 11y agoThere's still the right and secure way to go: 1) By default the cat is "locked", and uses a manufacturer-issued public key to verify the signature of updates 2) A car owner can "unlock" the car, by creating a key pair, and asking the manufacturer to add / replace (debatable) the new public key to the car. Right after that, the owner has an ability to modify the firmware, but everything is still secure. And it's up to the adult owners to manage their private keys and be careful with updates. 3) There should be a simple way to factory-reset the car (with either the owner, or the manufacturer signature).
- teraflop 11y agoUnder this model, if you buy a used car, how can you be sure it doesn't have keys that are controlled by someone else? You can follow the "factory-reset" procedure, but how do you know the reset mechanism hasn't been tampered with?
- mynameisvlad 11y agoI would assume in this perfect scenario that the factory reset lives in some ROM somewhere that nobody, not even the manufacturer can touch post-release.
- tzs 11y agoThey can design it so that the firmware is in two parts. The first part does minimal initialization and then gives control to the second part, which is where most of the functionality is. The first part is also the part responsible for implementing the firmware update protocol and the reset procedure. The first part would either not allow any update to replace the first part, or it would only allow the first part to replace with factory signed firmware. Alternatively (or in addition) they could have a diagnostic connector that lets external hardware read the firmware memory. You could then do the factory reset on your used car, and then hook something up to that diagnostic connector and have it compute a hash of the firmware that you can check to make sure it is the right firmware. An Arduino should be sufficient as the thing to read and hash the firmware.
- peterwwillis 11y agoWe (consumers) had to deal with that over 10 years ago. My old Lexus still has some kind of proprietary diagnostic system that only dealerships can access, and we get stuck with dinky ODB2. Did manufacturers suddenly start getting more transparent and compatible with their vehicle diagnostic and control systems? Also: virtually every consumer in the world would choose safety from some 'evil hacker force' rather than the ability to maintain their factory entertainment system themselves. There's really zero reason the nav/entertainment system should be on the same communications network as the engine and brakes, anyway.
- snowwrestler 11y agoPut a key on a piece of hardware and provide it with the car. The firmware will not update without that key. Now the owner can update their firmware, or hand the key to their mechanic to do it for them. If they lose the key, they get to pay their mechanic to source and install a new controller that comes with a new key. Or heck, just put a switch on the controller that must be manually closed for a firmware update to proceed. What matters most is preventing quiet firmware updates over the air.
- Too 11y agoEven if the update is not quiet, how do you know that it's authentic? If the hackers can trigger a "update available, please flip the update switch and restart your car"- notification, you will not be able to tell if this came from the manufacturer or the hacker.
- snowwrestler 11y agoIn my first idea, the hardware "key" holds two cryptographic keys. One is a private key, whose public counterpart is held by the manufacturer. This authenticates the vehicle to the manufacturer. The other key is a public key, whose private counterpart is also at the manufacturer. This authenticates the manufacturer to the vehicle. My second idea does not accomplish any new authentication (beyond what over the air updates already do). However it does prevent silent updates, which is an improvement over today. Today, not only can a vehicle owner not tell if an update came from the manufacturer or hacker, they can't even tell that an update happened at all.
- Qantourisc 11y agoYou could for example hook up the radio to a CANbus that has a minifirewall/proxy in it, allowing only read from the radio and write/read to the radio.
- PhasmaFelis 11y agoI had thought the firmware already required physical presence to alter? As per the recall notice, Jeep is mailing everybody thumbdrives with updated firmware. That suggests that the hack doesn't require changing the firmware, but merely exploits what's already there.
- mmozeiko 11y agoCharlie and Chris modified V850 firmware that is connected to CAN bus to control devices on CAN bus. They flashed V850 firmware from main info/entertainment system that runs Linux on ARM board. Chrysler most likely is patching this main firmware. Linux on this ARM board was entry point for this attack - it is connected to WiFi, 3G and other stuff and it has/had dbus listening on 6667 port for all interfaces.
- maxerickson 11y agoThey go into detail about how they flashed the controller in the paper, page 64.
- pacquiao882 11y agoThey were able to remotely flash the firmware while the vehicle is being driven down the highway and there's not really anything the passengers can do while it's happening. They are able to control other components like the audio system without needing to flash anything.
- dsuth 11y agoRegardless of the firmware question with the V850 module, I'm surprised that other modules on the CAN bus would accept commands from an entity that 'should' only have read access, without validation. I'm not up to speed on CAN bus, but if it's not possible to secure it in this manner, then it cannot be considered an appropriate communications bus for safety-critical applications. Even the industrial guys are getting their heads around this with encrypted & validated signalling over ethernet.
- mzs 11y agoCAN is a bus over a wire with no concept of source (or destination).
- kevin_thibedeau 11y agoCAN isn't really designed for security. The nodes don't even have addresses. They communicate using message IDs without regard to who the sender or responder is. The error detection requires a party line so that any node can flag a corrupt message and the data rate is too slow to make tunneling a secure stream practical.
- dsuth 11y agoThanks for this description (and others below). In that case it's not reasonable to use this for safety critical applications, IMO. Any element on the bus becomes a potential point of entry, and there's no defence in depth once on the bus. Kinda scary from an architectural point of view.
- billforsternz 11y agoUnfortunately you are casually proposing to trash untold billions, probably trillions of dollars worth of existing automotive engineering infrastructure with that comment. As any system is decomposed into components you can and must reach a level where inter-component communication is dumb, unquestioning, immediate. Obviously by the time you reach transistors (or pistons) that's how it works. CAN is not as fundamental to cars as pistons (or wheels if you prefer electric cars I suppose) but it's not that far removed. It was developed decades ago, at a time when this kind of hacking was strictly science fiction. I don't think it's necessary to throw away everything (as near as makes no difference) to address this problem. Real air gaps, real read-only circuitry etc. (as described in other comments) could do the job.
- sitkack 11y agoencryption is not the answer, physical read only circuits are the answer.
- ChuckMcM 11y agoThat is an answer, except consider the situation where a flaw is detected in the manufacturer's code. The cost to fix is to replace the entire chip. Which is fine, from a process standpoint, but these cars are going to be competing with other cars in the market and will have a higher "service cost" and that higher cost will dissuade people from buying (because historically less than 5% of the market is willing to pay a higher price for something that is "more secure"). So while I agree with you, unless it were federally mandated for all cars, the ones that didn't do it would have lower cost of ownership and get more buyers.
- Animats 11y agoThe buyer pays about $600 extra for the "Internet-enabled" version of the car. It's an option; you don't have to buy it, and, after this, probably shouldn't. If you didn't buy it, software updates require a dealer visit. I have a 2007 Jeep Wrangler with no external interfaces, and it's been in three times for software updates for basic automotive functions, such as stability protection and engine restart. Cars go back for recalls all the time. Having to physically replace a part isn't a big deal. Dealers have a supply chain in place for obtaining parts from the manufacturer.
- PhantomGremlin 11y agoIt's an option; you don't have to buy it, and, after this, probably shouldn't. Unfortunately, at least for some cars, all these options are bundled together in packages. Unless you buy the base model, you'll get the Internet as part of a package of other features that you want. I think that nowadays the better answer would be to go in and physically rip out the cellular antenna. Easier said than done.
- Pyrodogg 11y agoI don't think they implied that the chip should be read only, you should still be able to have a flashable firmware of some sort. But instead of giving this chip which should be read-only regulated by software direct access to the bus, route it through a bit of extra on board circuitry which physically prevents it from writing data to the bus. This would enforce the readonly client in hardware, not firmware.
- forgottenpass 11y agoIf the firmware is SECURITY CRITICAL then the only way to change it can be through physical presence, loading encrypted and and signed firmware Physical access? So, hypothetically another car gets hacked, but this time there is no middleman in position to implement a mitigation like Sprint was in this case. How do you suggest a firmware rollout happens? Recalls? Mailing thumbdrives to end users? Over the last week I've seen the infosec community warmer to the idea of OTA updates and all that baggage that entails compared to the alternative ways to update car firmware. You're posting pretty authoritatively though, if you've got some analysis that the rest of us don't, I'd love to hear it.
- hueving 11y agoThe basic idea is that this wouldn't have happened in the first place if it required physical access.
- dlgeek 11y agoAnd if it was a buffer overflow that lead to an arbitrary RCE instead of a firmware-overwrite, and the fix was to upgrade the code to fix the buffer overflow?
- klibertp 11y agoWhat buffer overflow? Obviously the firmware is formally proven correct and proven not to include any of those? (I'm half joking here - I suspect that's not the case and won't be the case for a foreseeable future, but I'd be happy to learn I'm wrong here)
- vessenes 11y agoI have for some time felt that robot cars absolutely need a physical disconnect lever. Nice and big, break glass, and you're driving off the wire. But, this hack reminds me that there's no such thing in a modern car; the whole thing is networked. So, yeah, physical updates of firmware. Although, in the article, it's a subsystem's firmware, which physically protecting that is just not in the minds of engineers right now: think of all the ports you'd need. Even if the physical kill switch left me with just (unpowered) steering and brakes, I would have a vastly different perspective on the safety inherent.
- Qantourisc 11y agoIn a lot of cars, this break-the-glass is still your ignition/car key. Big problems start when you have a "break-by-wire" model.
- Too 11y agoIn lots of cars the key doesn't do anything when the vehicle is moving. In some cars you can't even turn it mechanically if you don't have gear in P or R, in other cars you can't even turn it at all because its just a button. There are very good reasons for this, the recent GM issue where the key fell out by vibrations which killed the power steering comes to mind, some people actually died because of this.
- __david__ 11y ago> "Of course its secure, you can only READ the CANbus, the software doesn't even HAVE a write capability!" That can mean two things—either the software didn't implement writing but the hardware is capable of it, or the hardware isn't capable at all. Sounds like the former in this case. It should have been the latter.