There are ways to measure lies in a survey. I'm not an expert, but when I took a statistics class it was explained, and as well as I remember:
- Instead of the 1.000 surveyed people separate a smaller group, say 100.
- Assume the last digit of their telephone numbers are uniformly distributed
- Ask them something along the lines of: "If you lied about file sharing or if the last digit of your phone number is an even number say YES, if not say NO"
If nobody lied you should get something close to 50% YES and 50% NO... but if the result is different you can assume it's because some of them lied (although you don't know which). ... Doing this kind of test you can approximate the percentage of people that would lie to the question.