Collaborative filtering techniques are successfully employed in recommender systems to assist users counter the information overload by making accurate personalized recommendations. However, such systems are shown to be at risk of attacks. Malicious users can deliberately insert biased profiles in favor/disfavor of chosen item(s). The presence of the biased profiles can violate the underlying principle of the recommender algorithm and affect the recommendations.
This paper proposes two metrics namely, Rating Deviation from Mean Bias (RDMB) and Compromised Item Deviation Analysis (CIDA) for identification of malicious profiles and compromised items, respectively. A framework is developed for investigating the effectiveness of the proposed metrics. Extensive evaluation on benchmark datasets has shown that the metrics due to their high Information Gain lead to more accurate detection of shilling profiles compared to the other state of the art metrics.

Density based Subspace Clustering algorithms have gained their importance owing to their ability to identify arbitrary shaped subspace clusters. Density-connected SUBspace CLUstering(SUBCLU) uses two input parameters namely epsilon and minpts whose values are same in all subspaces which leads to a significant loss to cluster quality. There are two important issues to be handled. Firstly, cluster densities vary in subspaces which refers to the phenomenon of density divergence. Secondly, the density of clusters within a subspace may vary due to the data characteristics which refers to the phenomenon of multi-density behavior. To handle these two issues of density divergence and multi-density behavior, the authors propose an efficient algorithm for generating subspace clusters by appropriately fixing the input parameter epsilon. The version1 of the proposed algorithm computes epsilon dynamically for each subspace based on the maximum spread of the data. To handle data that exhibits multi-density behavior, the algorithm is further refined and presented in version2. The initial value of epsilon is set to half of the value resulted in the version1 for a subspace and a small step value 'delta' is used for finalizing the epsilon separately for each cluster through step-wise refinement to form multiple higher dimensional subspace clusters. The proposed algorithm is implemented and tested on various bench-mark and synthetic datasets. It outperforms SUBCLU in terms of cluster quality and execution time.