[summm]

They already add cryptographic authentication to some CAN messages, so you can't change them. It is only a matter of time until they add encryption.

This is mostly a corporate problem of risk aversion in my opinion. Some department writes down a risk assessment with a list of miniscule risks, for example of some 3rd party app backend being hacked. Or just a headline "Tinkerer hacked his car to use with his home assistant" in the local press. This list circulates, and since nobody in the middle management wants to be responsible for anything, and there is no officially approved positive use case, draconian countermeasures are drafted and constructed one by one.

[ornornor]

> draconian countermeasures are drafted and constructed one by one.

Except when it’s about privacy or anything else we actually care about: then absolutely nothing is done because it would cost more than 0 to do anything.

[reactordev]

On the contrary, lots are being done about it, they have to update their terms of service…

[summm]

Depends. In some sense EU companies are quite afraid of the GDPR. Privacy is used in a twisted way in that argument: if any privacy relevant data is exposed to another party, and there is any incident down the line, they fear they could be made responsible. So they to block you as a user to access your own data.

Of course, if that privacy risk came from them storing and selling your data, they happily accept that, you are right in that regard.

[SoftTalker]

They had to add this stuff because it was possible to unlock and start a car by accessing the headlight socket.

https://www.thedrive.com/news/shadetree-hackers-are-stealing...

[ryandrake]

> Or just a headline "Tinkerer hacked his car to use with his home assistant" in the local press.

It's pretty sad that "User used their product in a novel way we didn't expect" is seen as a risk that must be mitigated.

[therealpygon]

I suspect the manufacturer probably cares less about what you do to your own car and hacking it, than they do about the potential for security compromise of their products on a broader scale, where they will then get blamed and sued for not having closed said loopholes. It is a no-win situation when it comes to fault assignment.

[cisophrene]

> It is only a matter of time until they add encryption.

I hope I won't be in one of those cars when the in-memory encryption key gets bit-flipped by the unfortunate cosmic ray.

[b3lvedere]

Proving that autopilot killed that poor old granny because of cosmic rays would be an interesting case study.

[cisophrene]

It actually happened with Toyota around 2010: they went into a settlement regarding an unintended acceleration issue because it was proven the code was terrible and a single bit-flip could cause the behaviour.

https://en.wikipedia.org/wiki/2009%E2%80%932011_Toyota_vehic...

[AlotOfReading]

Bit of context to this, it was demonstrated that it was a hypothetical possibility, but the issue couldn't be demonstrated in lab conditions. Stuck floormats, pedals, and confused drivers remain the only actual explanations for the real events behind the lawsuits.

[b3lvedere]

Very interesting read. Thank you for the link.

[cisophrene]

Another interesting case: a proven case of bit-flip that affected a voting machine in Belgium: https://www.independent.co.uk/news/science/subatomic-particl...

[b3lvedere]

“ A mobile phone with 500 kilobytes of memory might only have one error every 28 years, but a router farm, such as those used by Internet providers, with 25 gigabytes of memory could have one every 17 hours.”

I wonder if current AI devices have protection against this…

[formerly_proven]

It’s a fair assumption that most of these things are trickle-down effects of CMS/R155 and CRA combined with very high risk aversion on the company side. The less you expose, the lower the risk.