[cryptography] cryptanalysis process?

Mohammad Hosein mhtajik at gmail.com
Thu Dec 26 16:41:40 EST 2013


depends on how the algo designed to be used in a crypto system -- usually
but not always :
- simple exhaustive attacks studied , see if its even worth it to get more
serious , OR , is it in the criteria of the "process" to continue that path
and trim down the cost by increasing the positive chances within realistic
time/budget/technology
- more known attacks such as differentials and birthday or so will be
studied . a possibility to tackle the algo and/or the crypto system with
TMTO could be categorized here as well . refer to books like "elementary
cryptanalysis" to surf common attacks to algo
- if the algo is serious and crypto system has complex regulations ,
efforts to find bugs within "implementation" of the algo or the built of
crypto system is another approach -- math dudes do not consider this as an
"attack" , which in that abstract realm they are right
- reading the algo and efforts to identify useless or destructive portions
of the code whether by human eye or a standard automatic system is also an
approach . a mistake in placement of some SBox somewhere , with assumptions
that are not always correct , as an example , may help the attacker shorten
the complexity of the attack in time , memory or computation power required
.
- in reality , most of the attacks hurt the security using flaws in crypto
system , whether design of built , and not the also itself . for engineers
, this is a quick and usually fruitful try with high chance of "getting
plain data outa some encrypted channel the way makes bosses happy and sell
9-10 figure price tags at ISS" , again , math dudes are not much friendly
here , when Shamir says " *Cryptography* is typically *bypassed*, not
penetrated" he is kinda referring to this , or other methods like using
human stupidity and not a pure math problem in an algo ,
- NIST has published detailed regulation and specs how to examine an algo
against sets of modern attacks . take a look . "code" must pass these tests
to even consider a crypto algo nowadays . NATO and ASEAN also have stuff in
this regard -- that i dont know much about . all public afaik


finally , confronting a crypto is not a black magic . knowledge is always
built upon knowledge and on these expensive and very intellectual areas , a
simple handbook is next to impossible to author .
one morning your SIGINT alarms that a new signal , a new channel , a new
paper or a microfilm hidden inside someone's butt is bring to their
attention and its apparently CODE . people with previous experince on both
math and engineering angle of crypto cracking ( e.g HPC ) study the signal
and with various methods , subject to signal analysis , at some point ,
revelations will be made about the general concept of the CODE . now there
is a system analyst , an engineer and a math snob sit together see about
the similarities of the CODE with other systems they have broken in the
past . if no ? dunno , maybe they pray ? if yes , the rest is almost always
the same . usually the crypto echosystem has a flaw or offers an
opportunity to make bosses happy , if not , above mentioned process on algo
starts . when i was young , the most important stage was getting from "we
have a CODE" to that room where "math snob , the engineer and an analyst --
usually a telecom expert with deep DSP and system design skills" get
together . no offense to algo , by the way . not sure if this style still
applies well or what .

regards
-mh



On Thu, Dec 26, 2013 at 2:37 PM, Kevin <kevinsisco61784 at gmail.com> wrote:

> Maybe it's just me, but the "soup to nuts" cryptanalysis process is black
> magic.  So I am curious...does one start with side channel attacks?  Which
> attacks are tried on an algorithm first and how is that decided?
>
> --
> Kevin
>
> _______________________________________________
> cryptography mailing list
> cryptography at randombit.net
> http://lists.randombit.net/mailman/listinfo/cryptography
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.randombit.net/pipermail/cryptography/attachments/20131226/16c7bc89/attachment.html>


More information about the cryptography mailing list