CRIME Open Source TCPA validation

From: Shaun Savage (savages@private)
Date: Tue Jan 28 2003 - 11:34:29 PST

  • Next message: George Heuston: "CRIME FW: [Information_technology] Daily News 01/28/03"

    -----BEGIN PGP SIGNED MESSAGE-----
    Hash: SHA1
    
    The point about stupid security, any sql server on the net, is a good
    point.  There are three strikes against any one who was directly
    effected by slammer.
    1>  SQL server on net
    2>  using Microsoft
    3>  Not patching
    YOUR OUT!!!!
    
    - --------------------------------------------------------------------
    
    Dept. of Homeland Security site switches to Linux from Windows 2000
    http://newsforge.com/newsforge/03/01/27/1831240.shtml?tid=2
    This happened during Slammer not because of Slammer.  To change the OS
    of the federal goverment takes along time.  This show that Linux can be
    more secure the MSwindows.  It had to take a act of congress to get
    security for the US by using Linux.
    
    - --------------------------------------------------------------------
    
    There was an article about IBM releasing TCPA code for Linux
    http://www.research.ibm.com/gsal/tcpa/
    This just moves the trust issue from software to hardware. Nobody trust
    Microsoft because of the backdoors and poor design concepts.  That is
    why Linux/Opensource offers more freedom and security.  If you want to
    check the code you can.  Now if the chip makers add their own backdoor
    or other "goverment requested features" are we not back to the clipper
    chip:(?
    
    I have thought about how to validate the hardware design for open
    source.  First the design needs to opensource, this is done by
    publishing the HDL of the chip/circuit.  Next on initalization
    after/during reset a checksum of the serial test path(boundry scan) is
    created.  This checksum is the accessed through a register during
    operation. it is the compared with a given checksum in software.
    
    Now what is to prevent the vendor from just returning the right checksum??
    Now it gets hightech with a serial test path scanner and generator. This
    is only used to verify the chip contains no backdoor.
    
    By resetting the chip and the hashing the serial output that should be
    the same checksum as the register.  There must be other input vectors
    that is shifted in, do one operation, the check the result with the know
    checksum for that operation.  This prevents the vendor from creating a
    seperate working section and faking the serial shifting.
    
    A security problem is if the private keys storred in the chip can be
    shifted out.  To check the validity of the chip/desgin they must, but to
    keep them secuer they must not.  To solve this problem the serial shift
    clock is connected to the reset pin of the registers that hold sensitive
    data.  The sensitive registers can only be reset once at the beginning
    of the serial test or the serial sequance will be effected.
    
    By using this scheme the design of the chip is first validated for no
    backdoors,  then each batch can be tested that it is the same and the
    design.  Now when a user reads the hash number from a vendor they can be
    assured that there are no backdoors.
    
    
    Now the TCPA chip/circuit is validated and the rest of the system uses
    this base validation to keep itself validated.
    
    Shaun Savage
    
    - --
    savages@private
    GPG2003 = 68DB 57E8 702C 21D9 0AA5  2375 1EB1 6F82 858C 23AD
    GPG2002 = B527 8F72 BAFA D490 6B30  6885 9FA2 34E8 EA73 F975
    Public key at  http://www.savages.net/gpg/savages
    
    -----BEGIN PGP SIGNATURE-----
    Version: GnuPG v1.2.0 (GNU/Linux)
    Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
    
    iD8DBQE+NtsjHrFvgoWMI60RAvH5AKCwFqaotBBcGgGxF228CxIIvETcxgCcDVir
    IrioV95ACcXYYjVjvydaLcw=
    =/Yyk
    -----END PGP SIGNATURE-----
    



    This archive was generated by hypermail 2b30 : Tue Jan 28 2003 - 12:07:14 PST