From: Pierre Uszynski <pierre@shell.portal.com>
Date: Tue, 26 Jul 94 12:05:44 PDT
To: cypherpunks@toad.com
Subject: Re: CYPHERPUNKS TO THE RESCUE
Message-ID: <199407261851.LAA22277@jobe.shell.portal.com>
MIME-Version: 1.0
Content-Type: text/plain


Let's not go overboard!!! I hear things like "public/private key",
"battery backup", "the door unit transmits", "a keypad on both sides"

Hey! This is a garage door opener! You think you can add $20 of stuff
and still have a profit margin?

How about no keypad, no battery backup on the remote, just a (random)
secret key, no door unit transmitter, normal batteries in the remote
and door unit:

1 - Plug the remote into the door unit. The door unit writes a roughly
random number (electronic noise), into the remote, and remembers the
same. Units synchronize time, while they are at it.

2 - Take the remote for a drive. When you come back press the "open"
button.

3 - The remote send the current time (as per its 1 min (im)precise
real time clock), encrypted with the secret key. See later real
message because of replay attacks.

4 - The door unit decrypts and checks against time, time +1, or
time - 1. If valid, it opens the door, and synchronizes its time
to the time stamp.

POWER LOSS. This is a garage door opener. What do you do when your
garage door remote fails, you try again, then you use your key and
replace the **** battery (start again at step 1). In my area, power
outages are so rare that I wouldn't even need a battery on the door
unit (your milage may vary).

REPLAY ATTACK. To foil an attacker who would record the time stamp
and replay it within a minute: After a time stamp opens the door, a
bare time stamp will only work again after something like 2 minutes.
Instead, when you push the remote "open" several times within the same
protected period, you issue "timestamp, tag1", then "timestamp, tag2",
etc... say you are allowed 6 "opens" in the same protection period.
The door unit has to check against time - 1, time, time +1, and the
corresponding 6 tags. When a tagged stamp worked, all the previous ones
are disabled for the protection period, whether the door unit
received them or not (to avoid the reuse of messages that failed to
open the door.) If you push the remote "open" and it fails to open the
door, and you simply give up on the idea of opening the door, you are
susceptible to replay attack for something like 2 minutes... that's life.

"CLOSE" ACTION. Just close the door, no crypto.

Any problem with this SIMPLE solution? Considering there is no point
in spending zillions on the garage door opener if your windows do not
have alarms.

The reason some current units may not be secure is probably the
usual: the manufacturers are not interested in what the customers
don't want and the customers don't want what the manufacturers have
never told them could (or should) be done. And politicians are more
interested in legislative action than in suggesting their constituents
do not buy junk.

Pierre.
pierre@shell.portal.com

ObQuotesFromPreviousMessages:

Sandy Sandfort <sandfort@crl.com> said something like:

> On Tue, 26 Jul 1994, Arsen Ray Arachelian wrote:
> 
> > You'd need a clock on the garage controller....The garage opener
> > would receive a signal from the remote, issue a challenge code based on a
> > hash of the time/date + some random numbers.  The remote would encrypt this
> > hash with the owner's IDEA key and send back the response.
> 
> Am I missing something here?  Why would you need a clock?  What I had in 
> mind was something like:
> 
> 	1--The owner presses the "open" button on the remote.
> 	2--The remote sends an "ask me" signal to the door unit.
> 	3--The door unit transmits a random number in the clear.
> 	4--The remote encrypts and signs the random number using
> 	   its unique private key.
> 	5--The door unit decrypts and compares the numbers, using
> 	   the remotes public key.
> 	6--If the numbers match, the door opens.  QED.
> 
> Adjusting my flame retardant underwear,
> 
> 
>  S a n d y
> 
> P.S.  For most car and garage doors, relatively short (32 bit?) keys 
>       should be more than sufficient, I would think.
>  
> 
> 
> 
> 
>