Keep an eye on your stats for suspicious activity. Install a click tracker such as AdLogger [adlogger.org] that can prevent multiple clicks etc...

After reading several posts here claiming that it was a good idea to add Google Analytics to a site, I signed up for it and put the code on every page.

Leave it to Google.

That's essentially what I do, and it's my understanding that having Google Analytics on your site makes it a bit easier for them to verify who did what.

How can Google Analytics help you finding out where the invalid clicks are coming from?

It's because Google has access to your visitor data through its Analytics software. They can track visitors and match their IP from Analytics to the clicks recorded by AdSense.

thanx every one
but why adsense don't disable invalid clicks and protect publishers?

but why adsense don't disable invalid clicks and protect publishers?

They do identify and throw out invalid clicks--not to protect publishers, but to protect advertisers (their paying customers).

So of these people who get banned for no reason know to them, how many I wonder are google anylitics. Is this really a protection against losing your account to click fraud?

I'm very reluctant to hand over such valuable information.

if google banned your adsense account, perhaps the first thing you'd want to do is to submit your server logs.

google analytics gives them a lot of that same info.

So of these people who get banned for no reason know to them, how many I wonder are google anylitics. Is this really a protection against losing your account to click fraud?

I'm very reluctant to hand over such valuable information.

If someone is a victim of click fraud, and G doesn't believe s/he isn't responsible for the clicks, s/he can be dropped from the program, regardless of whether s/he submits logs or uses G's analytics package.

Ask yourself this question. How can one prove one is not responsible for clicking on one's own ads? Are logs proof? What do they prove?

Also read up on Byzantine failure in computer systems for an understanding of determining whether or not some set of systems can be trusted to act in some particular way.