Source code

[Paul Rubin]

Alex Martelli <aleax at aleax.it> writes:
> > You'll have to give your customers a sealed computer and no permission
> > to install additional software, or communicate with the internet, if
> > you want to reliably prevent them from analysing your program.
> 
> Incidentally, the latter option can be feasible in certain situations: some 
> key parts of your program's functionality might reside on a computer you
> control, accessible to the rest of the application (installed more 
> traditionally on users-controlled computers) only as a black box via the 
> net.  XML-RPC, SOAP, and other distributed-computing approaches
> such as Corba, make implementing this particularly easy these days.

It doesn't have to be via the net: you can literally deliver a sealed
computer to the customer, with the sealing enforced by contracts and
inspections (this is quite normal, e.g. some financial applications
are delivered as a rack-mount box that you install in your machine
room) or by tamper resistant hardware (for higher security
applications).  For VERY high security, of course you can't let
hostile parties touch the hardware.  Nonetheless, the pay TV industry
deploys millions of sealed computers (smart cards) every year to
customers, every single one of which is considered an attacker.  While
that industry had spectacular early security failures, these days it's
a fairly solved problem and they limit losses to acceptable levels.

It doesn't sound like the OP was asking for this type of solution though.