[Cryptography] [cryptography] Browser JS (client side) crypto FUD

Bill Stewart billstewart at pobox.com
Sat Aug 30 16:12:08 EDT 2014


> >     Well, no.  Implementing HTTPS:// is hard.  It is simply out of the cost
> >     range of about 99% of the websites [0].  Otherwise they would.
> > Modern CPUs have crypto accelerators (e.g. AES-NI). https is cheaper
> > than ever before.

The computationally expensive part of HTTPS isn't the symmetric crypto,
it's the public-key signature step at the beginning (especially with PFS,
which makes it hard for the server to cheat by reusing the same
session key for a bunch of sessions, if anybody still does that.)
DES benefited from hardware acceleration because it bit-twiddled
in ways that then-current CPUs weren't good at, but RC4 didn't need it.

>The thing that upsets mass rollout of HTTPS is the configuration,
>certificate, IP# and associated sysadm costs.  Only big-end merchants
>grumble about the CPU costs, but they have always been able to afford it.

Pretty much true. The original problem was time-delay in session initiation,
which used to be long enough that website metrics people worried about
customer satisfaction and abandoned visits.  CPUs are a lot faster
and have vector processing (SSE,etc.) that can speed that up.

On the other hand, HTTPS-everywhere also interferes with load-balancing,
because you can't just randomly assign pages to a lightly-loaded server;
sessions have to be stickier.

>So is 40 bit SSL which would have stopped PM in 1995.  But the world
>said it wasn't good enough so we fought the NSA for 128 bit, nothing but
>128 bit dammit!!!  Look what we got...
>
>If cypherpunks had had the smarts to shut the fluff up and let 40 bit
>pervade the planet, then we'd have upgraded to 64 bit then 80 bit then
>128 bit by now and the job would be done.

I'm assuming this is humor whooshing over my head?
40-bit RC4 was the same speed as 128-bit RC4.  Web commerce needed
crypto that wasn't trivially broken, to protect credit card numbers,
and realistically, while we got crypto out there and demonstrated
that weak crypto was weak, it was the money and not the civil liberties
that got widespread adoption past the NSA/FBI.




>Citation needed ;)  Actually, you need that, otherwise it seems
>nonsense.  I don't know anyone using FireSheep, how can it have made a
>difference?  "Massive" ??

IIRC, from what I read on the net (:-) at the time it happened,
FireSheep was what prompted a number of major websites to adopt https.





More information about the cryptography mailing list