Fraud in science, at least "major" fraud, is very rare. Any kind of falsification of results is very, very wrong, but I can imagine it's possible to get away with a tweak or two to data to make it look better. Outright falsification though is virtually impossible to get away with. The reason for that is that science is built around an extensive system of checks and balances. Important experiments will be repeated by scientists other than the one who did the original work. Although we scientists are a trusting bunch, we do like to make sure we can get relevant experiments to work in our own labs. If you can't get an experiment to work, you contact the original scientist and try to work out why. If no one can repeat the work, then falsification begins to look like a real possibility... If you do important science, someone will check up on your work. The more important/high profile your science, the more people will attempt to replicate it. Sir Peter Medawar, in his classic book "Advice to a Young Scientist" advocates that all scientists should work on something important. Some important science may seem esoteric to the public, such as perhaps my work on protein structure. Other science has spent some time in the spotlight, such as stem cell research, and is generally considered important in one way or another. I suspect most scientists believe what they're doing is important in some sense.
So one wonders, WHAT WAS HE THINKING????