Collaborative filtering-based recommendation systems algorithm are liable to shilling attacks beacuase these systems are open in nature. Shilling assaults are when malicious people give the target item the maximum/minimum rating in an effort to raise or lower the rating of the target item. These attacks effect the accuracy of recommendation system. For improving the accuracy of recommendation system we introduced a new feature i.e. Skew Deviation Bias which measures variable's distribution asymmetry. In this paper amazon product data is used as dataset and the supervised learning algorithm i.e K nearest neighbour is applied for detection of shilling attackers. This new feature increases the accuracy of detection of shilling attackers using KNN algorithm by 2 to 5%. Tests are being done on amazon dataset and comparison is also being shown between KNN with existing feature set and KNN with new attribute.

K nearest neighbour ; Skewed Deviation Bias; Shilling attack; collaborative filtering