TCB

Trusted Computing Base

remark this page should probably be removed in favor of "reliance set", which would include human operators as well. See http://www.erights.org/ Mark Samuel Miller thesis pg. 29.

A TCB consists of a set of hardware and software mechanism which guarantee that security will not be violated under any circumstances.

A TCB is must be formally proven to be correctly implemented. (Paying attention to the Trusting Trust issue.)

The /. article of 18th Aug 2004 introduces http://www.rpow.net/ a second TCB based on a slightly different approach.

See also http://www.dtcp.com/

Current approaches try to provide a cryptographic check sum (in hardware), which is updated after a certain amount of processing has been done. (E.g., after the boot loader has been executed, after the operating system has been loaded etc. This is also how thecurrent BALL implementation computes the opaque part of the version slot of a place.

open tc aims to provide a free implementation, focusing on linux.

As of 2003/4 there are concerns about bad designs for secure hardware under the names TCPA and TCG. Read more on http://www.eff.org/Infra/trusted_computing/20031001_tc.php An good faq on the usual problems accociated with central control http://www.cl.cam.ac.uk/~rja14/tcpa-faq.html especiall interesting the last question concerning the DoD definition of trust: "a trusted system is one, which can break the security policy" - which is correct in so far, as the the ability to break the policy is what requires client to trust the system not to do so. For "anonymity" - which means the guarantee not to let untrusted parties to spy on secrets - this definition is correct. Integrity however is a different thing, there is no such thing like a correct system, which can break the correctness property aka. "security policy". To avoid confusion we better don't call such systems "trusted". But what else? Correct? http://www.cypherpunks.to/TCPA_DEFCON_10.pdf

The main concern is that users are not be in 100% control of their own environment. (This groklaw article illustrates the problem from a legal background. I'm not yet sure that this could happen on the platform.) If such a system would run after a some non-obvious software was loaded "in the name" of a user (e.g., if it was possible to load an encrypted bios or system kernel) than that user was already impersonated. Since the structure of those plattforms is equivalent to "botnets", they are highly vulnerable to abuse nnd therefore basically useless for lawful purposes. Once ebmraced it could no longer incure legally binding consequences.

some critical voices: http://www.protectprivacy.org/ , http://www.notcpa.org/

The german computer magazine publisher heise reports chaos computer club critism on the dangers of "trusted computing" http://www.heise.de/newsticker/meldung/54655 .

---

There are more computer related uses of the acronym: see folddoc. The one I like most: Trouble Came Back