3 ms·

What I really want is a hardware switch that disables any kind of input from the "self-driving" software. Something that cannot be overridden by an OTA.

buran77 7y ago
The system is pretty deeply integrated with the rest of the car. So I imagine turning it off would take more than one single "flip". Or at least turning it off reliably, since every module involved has to be disabled while the rest of the car acts normally and is still connected to the internet.
- java-man 7y ago
  This is, in my opinion, rather problematic from a security point of view.
m463 7y ago
First, the self-driving features are all opt-in. Also, a lot of features are passive, like sound alerts. You have to enable things like Autosteer, Summon or Navigate on Autopilot in settings before you can use them. Then you have to hit a switch while driving to let the car help you drive. As for emergency features (which all cars get), they also have to be enabled in settings, like Lane Departure Warning, Automatic Emergency Braking and Object Aware Acceleration. My question to you: what if you turned off Object Aware Acceleration and then ran into something... or someone?
- java-man 7y ago
  I am looking at it from a security perspective. Any sufficiently complex software system, especially one that does not have security as prime design goal, exhibits a large attack surface. It is not a matter of if, but a matter of when. Example scenario: a CEO of a large public corporation (which attracted a lot of short sellers). A journalist who have exposed some far-reaching shenanigans of the rich. A political opponent. I am sure all the relevant agencies are quite busy analyzing Tesla code right now, for exactly this reason. And I know for sure that security is NOT the design goal at Tesla, at least when I talked to them. I would have designed the hardware differently from the start. Again, I am not talking about a common use scenario for regular people, or driver turning off some safety feature. I am concerned with security of the system as the design goal.

buran77 7y ago

The system is pretty deeply integrated with the rest of the car. So I imagine turning it off would take more than one single "flip". Or at least turning it off reliably, since every module involved has to be disabled while the rest of the car acts normally and is still connected to the internet.

java-man 7y ago
This is, in my opinion, rather problematic from a security point of view.

m463 7y ago

First, the self-driving features are all opt-in. Also, a lot of features are passive, like sound alerts. You have to enable things like Autosteer, Summon or Navigate on Autopilot in settings before you can use them. Then you have to hit a switch while driving to let the car help you drive. As for emergency features (which all cars get), they also have to be enabled in settings, like Lane Departure Warning, Automatic Emergency Braking and Object Aware Acceleration. My question to you: what if you turned off Object Aware Acceleration and then ran into something... or someone?

java-man 7y ago
I am looking at it from a security perspective. Any sufficiently complex software system, especially one that does not have security as prime design goal, exhibits a large attack surface. It is not a matter of if, but a matter of when. Example scenario: a CEO of a large public corporation (which attracted a lot of short sellers). A journalist who have exposed some far-reaching shenanigans of the rich. A political opponent. I am sure all the relevant agencies are quite busy analyzing Tesla code right now, for exactly this reason. And I know for sure that security is NOT the design goal at Tesla, at least when I talked to them. I would have designed the hardware differently from the start. Again, I am not talking about a common use scenario for regular people, or driver turning off some safety feature. I am concerned with security of the system as the design goal.