Re: A89: Linux Port for 89/92

[Prev][Next][Index][Thread]

Re: A89: Linux Port for 89/92


Ok, let's put it this way:

We have, or can have, full and complete access to the contents of the flash rom,
where the boot loader is, I am told, located.

It would be trivial to make a program that would run tests on that code within a
certian range, and have everyone who owns a ti-89 who uses ticalc.org, these
mailing lists, and the newsgroup bit.listserv.calc-ti run the program on various
ranges.  I think we could amass a large enough amount of computing power to give
us a good chance at breaking it soon.

Now, I completely understand that it still a matter of chance that we could find
it in under 5 years (unless we could get one of the larger internet distributed
processing sites on our side...), but that chance is much *much* greater than
finding it by having us sitting around shooting other people's ideas down
without really thinking about them. Hmm?

But it's really irrelevent, because we can tack the same encrypted checksum on
the end of whatever code we introduce, and add a few bytes as necessary to match
the checksum it will be checking for.

Secondly, we'll have a second checksum at the end of this year.  Should narrow
the field of search.


-Adam


Johan wrote:
> 
> Adam Davis wrote:
> >
> > It is a trivial matter to brute force break the code.  We just need someone
> > willing to hook the calc to their computer.
> 
> While the .tib file is being downloaded, the boot loader calculates a
> checksum of the code. At the end of the .tib file, there is an RSA encrypted
> version of the same checksum. The boot loader decrypts the checksum and
> verifies that the calculated one equals to the decrypted one and refuses to
> run the code if they don't match.
> 
> The key that is required for *decrypting* is available to us, but we can't
> *encrypt* our own checksum.
> 
> The RSA key that TI uses is 512 bits and the MD5 checksum is 128 bits.
> Trying to adapt a ROM to get a specific checksum is would take (on average)
> 10^25 years if the computer can do one million tests per second. Breaking
> the RSA would take 10^140 years for the same computer.
> 
> If you still think it's a "trivial matter", please tell me and the rest of
> the list how you are going to deal with it!
> 
> //Johan
Follow-Ups: References: