Comment Re:Adversarial Implications of sharing information (Score 1) 140
That problem is solved if cars act as if all the information they can trust is their own, and only add "potential dangerous situations" reported by others to their own list, but never discarding them purely based on another machine's information.
This approach has the same issues that we have in cyber security (i.e. think x509 certificates and Certificate Authorities). How do you know who to trust? Can we always trust them? When should we trust them? If we use a reputation system to manage trust, how do we make it work such that it scales?