[41361] in Cypherpunks
Re: responce to graphic encryption replies
daemon@ATHENA.MIT.EDU (Futplex)
Fri Oct 13 00:36:29 1995
To: privsoft@ix.netcom.com (Steve Orrin)
Date: Fri, 13 Oct 1995 00:32:51 -0400 (EDT)
Cc: cypherpunks@toad.com (Cypherpunks Mailing List)
Reply-To: cypherpunks@toad.com (Cypherpunks Mailing List)
In-Reply-To: <199510121733.KAA18977@ix7.ix.netcom.com> from "Steve.Orrin@ix.netcom.com, Mgr.Tech.Services@ix.netcom.com," at Oct 12, 95 10:33:35 am
From: futplex@pseudonym.com (Futplex)
Steve Orrin writes:
> Also, I have recently put together an info sheet on the Security provided
> by PrivaSoft which I can post if there is interest.
I for one am interested. Perhaps you could put it up on your web pages ?
[...]
> One of the key strengths, as I see it, of graphic encryption is
> that during decryption via hacking, there is an added time element when
> a human interface is required to verify the product, ( since it is a
> graphic picture being produced, regular checksums for intelligible
> words can't be used sans implementing OCR), even if this is only 10
> milliseconds per try this is increases the time to crack
This is an interesting point I hadn't previously considered. Can anyone
comment on the state of the art in fast approximate character recognition ?
I expect that the people working on recognition of text in TV pictures etc.
would have a good idea.
My lay computer scientist's guess is that it wouldn't be all that difficult to
pick a small sample window a couple of characters wide, and decide if the
contents were a couple of characters. Then you'd worry about testing for
higher-level linguistic intelligibility as a second cut. But I don't
really know.
A known-plaintext attack on the system would ideally include knowledge of the
typefaces, fonts etc. typically used to print documents at the source....
> exponentially
> beyond that of a data encrypted document of similar key length and
> algorithm strength.
ObTheoretician:
Um, exponentially in terms of what ? It sounds like this multiplies the
expected brute force cracking time by a constant, but doesn't change the
big-O time of the algorithm. I agree, however, that big constants can be
rather significant when it comes to real world applications.
-Futplex <futplex@pseudonym.com>
