My 2 cents on the subject...

The automatic toll fee system I am most familiar with is that of Kapsh (used
to be Combitech).  They have implemented automatic toll fee collection in
many countries around the world (in Europe, Asia, Australia, south

I think they usually implement a combination of 
1) a system that queries a device in the car, which identifies the car
owner, and then charges the owner in a central database (incrementing the
amount that is due)
2) license plate scanning for accountability purposes.

When you do crypto to authenticate the communication between the toll device
and the device in the car, you need to do fast crypto.  Where I work, we
used to be in the hardware arena and had a project designing an HSM for a
toll fee system.  The requirements where that it had to be based on DES/3DES
and you had to be able to do DES/3DES operations on single, small length
messages, rapidly.  This last part is a bit tricky, it's not the same as
getting good average speed on longer messages, you need to take into account
the communication between the PC and the HSM which accounts for allot of
overhead on a single, small length message;  IO memory mapping is a good way
to go, also preparing keys in RAM can help just a bit, but for us IO memory
mapping gave the most significant speed-up.  There a paper from IBM on this
subject (can't find the reference now), with the same conclusions.

License plate scanners seem to be effective these days.  I related story to
the toll fee license plate scanning, Toronto police are using a license
plate recognition device to scan parked cars in order to attempt to identify
stolen cars:

They were able to recover 153 stolen cars in a 3-month test period.
They say they can scan 1000 license plates an hour, but this includes the
time to send the information to a central point and do a search in a


The Cryptography Mailing List
Unsubscribe by sending "unsubscribe cryptography" to [EMAIL PROTECTED]

Reply via email to