[Cryptography] Crypto for optimistic transactions ?

[Tom Mitchell]

On Thu, Jan 4, 2018 at 8:55 AM, Henry Baker <hbaker1 at pipeline.com> wrote:

> As I pointed out in my previous posting, there is no "undo" for learning a
> bit of information, so it is *impossible* to truly "roll back" an
> optimistic transaction if that bit is inadvertently disclosed during the
> negotiation.
>
> So is there a role for crypto in solving this problem?


For hardware the Ben Franklin analysis might come to play.
Hardware can kill all the bits that hold speculative information.
Alternatively they need never be alive.

    "Three can keep a secret, if two of them are dead."  << == killing
hardware registers is possible.

The race condition eliminated and timing of the logic sorted the Ben F.
solution can apply
Already logic blocks allow out of order execution but the result is not
enstantiated
until it is time.   These extra result registers allow hyper-threading.

The first problem is what to do now.  I suspect the goal is to slam the
door on the risk.
Intel will be happy to make AMD collateral damage in the short term.  The
urgency of the Intel
mess cleanup will dominate and so far the Intel driven solutions do not
seem to break AMD
just slow both down.

A second is to address the reality that there are decades of hardware to
make safe (enough).
Decades of hardware is not going away.   A personal machine running known
software is not
a worry where the owner of the machine has administrative rights and sudo
access.

The big risk is all the hardware that makes the cloud easy to manage and
make the cloud possible.
Once the door is slammed on this some sanity will prevail.

I fear that this attack set will spawn more attack research and more chinks
will be found.

The crypto risk is that data systems all depend on credentials and should
these leak
all manner of problems follow.

A third landmark goal is to design hardware that does not take the 5-35%
perf hit.  Whoever solves this
first will sell a lot of hardware including motherboard and DRAM makers.
 But this will put a chill on
sales for months.






-- 
  T o m    M i t c h e l l
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.metzdowd.com/pipermail/cryptography/attachments/20180104/8a54749c/attachment.html>