Filtering "fake or deceitful reviews" is a complicated task to accomplish. While you won't find the exact solution, you can research how companies like Yelp and AngiesList have done it.
There are many factors in the formula, but a lot of it revolves around looking at who the user is exactly. By looking at how often they contribute, how long they have been a member of the site, and incorporating a rating system into the review (helpful/not helpful) you can begin to determine which are the best reviewers.
Verifying the user is real is another step you can take. You can do this by adding a verification step into the account creation process (i.e. responding to a email) prior to submitting a review. Setting default time limits to how long the account has to be active before they can make a review also can work (this avoids the random user in the heat of the moment writing a false review or one that is not fully thought out). If your users have "profiles" in the system, similar to Yelp, you can give more credibility to a user who takes the time to fill out their profile info.
Another method is to track IP addresses and compare to where the place of business being reviewed is. Although this will not work in all cases (i.e. someone traveling or logging in from home for a vacation experience), it does lend some credibility.
By giving each method you incorporate into your system a weight, you can then add the weights up to come up a validity indicator. Each method could be weighted on a scale of 1 to 10 and the higher the weight awarded for that method the more credible. You then have to weight the methods in general. In the end the formula combines all those weights together to get a total weight which should give you an idea of how credible the review is.