[ Date Index ] [ Thread Index ] [ <= Previous by date / thread ] [ Next by date / thread => ]
On Fri, 6 Sep 2013, bad apple wrote:
So, to summarise: completely breaking our current crypto algorithms would be as spectacular as exceeding the speed of light or showing an instance where entropy did not increase in a finite system over time. It may well never happen. Incremental improvements in attack strategy, algorithms and spotting implementation attacks are however a given.
We agree. :-) The two reasons for my comments were:1. the distinction between the maths and the implementations isn't always clear. Many RSA implementations don't chose the prime numbers in a random enough way. That's an implementation issue, but may not look like it. I worry people confuse the two and think "well, I'm using 4096-bit RSA, what could possibly go wrong".
2. when crypto gets harder than RSA (elliptic curves and beyond) the chances of there being a flaw or vulnerability in the maths are bigger. The chance of there being something fundamentally wrong with the maths in ECC is more or less zero, but ECC involves choosing a certain curve. There are classes of curves that lead to weaked-than-expected crypto that you should thus avoid.
It possible that an entity knows other curves that are weak. Through clever social engineering, the entity may even be able to nudge people into using curves from this class for their ECC.
Reading comments from some experts, I don't think this is likely and I don't think this is what happened - Schneier's and Snowden's comments strenghten me in this belief. But it is a possibility that is not to be completely excluded.
Martijn.(Disclaimer: I did maths at uni and the did some research into algebraic geometry, which gave me some insight into how elliptic curves and ECC work. "Some insight" is nowhere near enough for cryptography though and I haven't done any of this stuff for more than seven years.)
-- The Mailing List for the Devon & Cornwall LUG http://mailman.dclug.org.uk/listinfo/list FAQ: http://www.dcglug.org.uk/listfaq