[Cryptography] randomness +- entropy

[Jerry Leichter]

On Nov 8, 2013, at 8:19 PM, Bill Stewart <bill.stewart at pobox.com> wrote:
> Most smartphones have wifi, and can tell you things like signal strength and noise levels on different channels, even if they don't show you the cellular signals.  It's probably not a rapidly-changing signal, but it's going to have some entropy (and if there's not enough noise, go microwave some coffee.)  There's also typically an accelerometer, and even dumb phones almost all have cameras.
> 
> But unfortunately, that $29 cable modem or dsl router isn't going to have any spare hardware, even a 5-cent photocell, so it's network noise only unless there's wifi.
Every DSL modem I've ever seen can display noise information in multiple frequency bands.  (It's always gathering this information to adjust its use of the link.)  Firmware would have access to it, and it's likely a very good source to use for driving an RNG.  (Yes, you'd want to first invest some effort in determining exactly how this data is sampled, whether it's exported anywhere - the adjustment is done by the two ends of the link together, but I know nothing about what specific information is exchanged between them, etc.)

I know nothing about cable modems, but most likely they have access to similar kinds of information:  Sending data across long, uncontrolled spans of wire will generally require some sort of adaptation to the characteristics of that wire.

All that said ... I have yet to see a DSL or cable modem that *needs* a secure source of random numbers.  They live at L2 and below and don't encrypt or decrypt anything.  OK, they usually have http interfaces for management - which should really be https and they should come pre-configured with a certificate, as they come with a unique password.
                                                        -- Jerry