Building a High-Performance, Programmable Secure Coprocessor

Replaces RC21045

Secure coprocessors enable secure distributed applications by
providing safe havens where an application program can execute (and
accumulate state), free of observation and interference by an
adversary with direct physical access to the device.  However, for
these coprocessors to be effective, participants in such applications
must be able to verify that they are interacting with an authentic
program on an authentic, untampered device.  Furthermore, secure
coprocessors that support {\em general-purpose} computation and will
be manufactured and distributed as {\em commercial products} must
provide these core sanctuary and authentication properties while also
meeting many additional challenges, including:

  * the applications, operating system, and underlying security
 management may all come from different, mutually suspicious
 authorities;

  * configuration and maintenance must occur in a hostile environment,
 while minimizing disruption of operations;

  * the device must be able to recover from the vulnerabilities that
 inevitably emerge in complex software;

  * hardware constraints dictate that support for advanced cryptography
 depends on reloadable software; and

  * physical security dictates that the device itself can never be
 opened and examined.

This paper summarizes the hardware, software, and cryptographic
architecture we developed to address these problems.  Furthermore,
with our colleagues, we have implemented this solution, now available
as a commercial product.