What about encrypting the data (text files), reading the encrypted data at runtime, decrypting in memory, and then assigning to the variables?  (Yes, someone skilled enough could extract the decrypted data from RAM, but let's assume that's a necessary casualty.)

On Wed, Dec 7, 2011, at 12:19 AM, Ian Kilgore wrote:
> On Tue, Dec 6, 2011 at 10:12 PM, Daniel Underwood
> < daniel.underwood@ncsu.[redacted] > wrote:
> > The point of the program is to generate a sample of a random variable with distribution defined by the input data I'm trying to conceal from the user.
>
> The best you can do is make it kind of annoying for a skilled user to
> recover the data, period.
> Unless you make it a client-server thing, which is rarely practical
> and usually slightly more annoying.
> --
> ik
>

--
Daniel Underwood
North Carolina State University
PhD Student - Industrial Engineering
email: daniel.underwood@ncsu.[redacted]
phone: XXX.302.3291
fax: XXX.515.5281
web: http://www4.ncsu.edu/~djunderw/