Smart meters are deployed in massive numbers and maintenance/upkeep of flaky units is an intractable problem. They are BOM-constrained products. So they tend to use dirt-simple bulletproof embedded processors; mostly, MSP430.
So provisioning "2048 bit keys" (more realistically/reasonably: aggressively small ECC keys) is not as simple as pulling them out of the air and assigning them.
On top of that, and for the same reasons I mentioned above, smart meters tend to communicate over very simple RF protocols. There isn't a lot of headroom in those protocols for large moduli or, for that matter, for many round-trips.
I'm not arguing that these systems should be more secure. They obviously should be. I'm disputing that securing them is an easy problem. It is not.
Later
(There are better smart meters, as someone mentioned downthread. But remember also that it is very, very difficult to get cryptography right. It has essentially never been done in the first major deployment of any system.)
Agreed! Securing stuff is NOT an easy problem--especially cost constrained, embedded goods. Having worked on multiple connected device projects now (including Square's first encrypted reader--but that was one of the more rigorous efforts I've seen), I'd say there's not much communication in our industry & everyone seems to roll their own security. It's disastrous & I'm not certain what the solution is. Only companies with obvious liability issues seem to bother with security consultants such as Matasano. The rest wing it with reference designs and code scraped from dubious places. So, when articles like this surface, I'm fascinated, but not surprised. BTW, this problem is only going to get worse with the IoT boom.
I have nothing but respect for what you do, Thomas. I wish everyone I've worked with had access to your services. Could you maybe help TI wrap it all up into a hardware block in the next generation of MSP430s? :)
So provisioning "2048 bit keys" (more realistically/reasonably: aggressively small ECC keys) is not as simple as pulling them out of the air and assigning them.
On top of that, and for the same reasons I mentioned above, smart meters tend to communicate over very simple RF protocols. There isn't a lot of headroom in those protocols for large moduli or, for that matter, for many round-trips.
I'm not arguing that these systems should be more secure. They obviously should be. I'm disputing that securing them is an easy problem. It is not.
Later
(There are better smart meters, as someone mentioned downthread. But remember also that it is very, very difficult to get cryptography right. It has essentially never been done in the first major deployment of any system.)