[linux-elitists] defanging TCPA?

Don Marti dmarti@zgp.org
Mon Feb 3 10:33:38 PST 2003


TCPA does not seem to be particularly dangerous to freedom or
effective for mass-market DRM as long as (1) many computers have
different kernels, and (2) nobody but the computer's owner knows
bit-for-bit what's in that computer's kernel.

1. When installing an OS or first booting a new computer,
people should run a program that changes the content of the
on-disk kernel image without affecting its functionality,
by replacing "synonymous" instructions.  One way to do this
is generate a random string and hide it in the kernel with
something like Hydan, which is steganography for executables:
http://codecon.info/2003/program.html#hydan

2. Point-and-click installers for commonly downloaded programs,
both free and non-free, should check the filesystem for kernel
images, compute some hash function of each one, and either refuse
to install or issue a warning if the hash matches a well-known one.

The big problem with pursuing kernel uniqueness is -- what about
recovering the contents of my personal data on my personal encrypted
partition whose key is stored using TCPA?  But if your computer
fails or you upgrade it, and you move the disk containing that
partition to another machine, you'll still need a non-TCPA way of
getting at the key, such as typing in a passphrase.  You can do that.

-- 
Don Marti                  Even if we don't get DMCA reform, loudly
http://zgp.org/~dmarti     demanding DMCA reform is going to get the
dmarti@zgp.org             injustice of the DMCA in front of the next
KG6INA                     jury.  Make noise.  It counts.



More information about the linux-elitists mailing list