No abstract

No abstract

Abstract. Role Based Access Control (RBAC) has been introduced in an effort to facilitate authorization in database systems. It introduces roles as a new layer in between users and permissions. This not only provides a well maintained access granting mechanism, but also alleviates the burden to manage multiple users. While providing comprehensive access control, current RBAC models and systems do not take into consideration the possible risks that can be incurred with role misuse. In distributed environments a large number of users are a very common case, and a considerable number of them are first time users. This fact magnifies the need to measure risk before and after granting an access. We investigate the means of managing risks in RBAC employed distributed environments and introduce a probability based novel risk model. Based on each role, we use information about user credentials, current user queries, role history log and expected utility to calculate the overall risk. By executing data mining on query logs, our scheme generates normal query clusters. It then assigns different risk levels to individual queries, depending on how far they are from the normal clusters. We employ three types of granularity to represent queries in our architecture. We present experimental results on real data sets and compare the performances of the three granularity levels.

This study presents the design and implementation of DES encryption in ECB (Electronic Codebook) mode under three different parallelization schemes. The first method we suggest is block based, where 64-bit blocks of text are distributed among the parallel processors. The second approach is pipeline based, where we parallelize each of the 16 internal rounds of DES. The third parallelization scheme is plaintext based which divides the plaintext into equal partitions first and then assigns the 64-bit blocks of each partition to parallel processors. We obtained runtime performances for each parallelization. To observe whether the source language has an effect on the overall performance, we run our algorithm on text from 5 different languages as English, French, German, Spanish and Turkish.

No abstract