RESEARCH GROUPS ACTIVITIES

CRDC at BZU is directly related and actively linked to the sector of Science and Technology. Our people include a proliferant blend of accomplished faculty members and each one is actively involved in research activites. Some of these are given below...

ON THE ANALYSIS AND IMPROVEMENT OF DISTRIBUTED SORTING ALGORITHMS


By:

          Farhad Mubashir [MS-10-01]

Supervised By:

         Dr. Minhaj Ahmad Khan

This research explores the performance of different distribution and comparison sorting algorithms designs, implemented in distributed environment, by exploring their performances. The issue is addressed by finding the best sorting implementation on a particular architecture for a specific slot (range) of values in a distributed environment. To accomplish this task, an algorithm is proposed which generates a decision table with actions representing the implementation of most appropriate sorting algorithms.

DYNAMIC CODE GENERATION & OPTIMIZATION FOR EXPLOITING PARALLELISM


By:

          Adeel Iqbal [MS-10-04]

Supervised By:

         Dr. Minhaj Ahmad Khan

In this research, Hybrid Java Parallelizer (HJP) framework is proposed and a prototype of HJP is designed as well. HJP framework can convert the possible parallel portion of the code in given module from sequential to parallel automatically without any directives and there is no need of source code as it can convert / decompile class file to source by itself and use it for further processing and parallelizing the code. HJP’s distinct feature is that it can create the number of threads according to the machine specifications on which it is executed. This means that HJP has the ability to properly utilize the available resources to execute huge computational tasks easily and efficiently in minimum time.

EFFICIENT COMPRESSION ON HETEROGENEOUS SYSTEMS


By:

          Muhammad Ali [MS-10-10]

Supervised By:

         Dr. Minhaj Ahmad Khan

In this research, an algorithm is suggested for performing XML compression and decompression. The suggested approach reads an XML file, removes tags, divides the XML file into different parts and then compresses each different part on a separate core for achieving efficiency. Performance results are then compared with other algorithms on different architectures to find notable performance leap of our proposed algorithm.

EFFICIENT REMOTE METHOD INVOCATION (RMI) FOR MATHEMATICAL COMPUTATION


By:

          Muizzud-din [MS-10-13]

Supervised By:

         Dr. Minhaj Ahmad Khan

Remote Method Invocation (RMI) is the modern way of Java to encompass with the concept of distribution. It is a client server application model which provides greater control to clients with a powerful server. The purpose of this research is to use the techniques which will improve the overall efficiency of a particular system of computation in terms of processing time. For this purpose, parallel programming capabilities, some compression libraries and RMI is used. Various algorithms are investigated and tested in this perspective. Analysis of different compression libraries is performed and effect of parallel programming using threads is observed.

COMPARATIVE ANALYSIS OF EFFICIENT PLATFORMS, ALGORITHMS AND PARALLEL PARADIGMS FOR BULK IMAGE PROCESSING


By:

          Khawaja Tehseen Ahmed [MS-08-16]

Supervised By:

          Dr. Qaiser Rasool

This research presents a comparative analysis of different programming languages and libraries to parallel process a huge stream of large and small images which undergo unpredictable arrival and time variation. For this parallelism, different levels of implementation with different algorithms and techniques have been investigated. Capabilities of existing tools for real time parallel processing has also been surveyed. Experimentation performed on hardware platform proficiencies for large data stream processing. It also focuses the state of the art discussion of various existing technical & architectural solutions to implement the parallelization on a hybrid platform for real time processing of large number of images in a stream. Moreover, experimental results obtained using Apache Hadoop in combination with OpenMP have also been discussed.

A SYSTEMATIC REVIEW OF SOFTWARE FAULT TOLERANCE


By:

          Farial Syed [MS-08-05]

Supervised By:

          Muzaffar Hameed

This research presents a study of fault types with their categorization agreement by the various fault classifiers, software fault tolerance. After a brief overview of the software faults, software fault tolerance techniques are reviewed. For some applications software safety is more important than reliability, and fault tolerance techniques used in those applications are aimed at preventing catastrophes. Single version software fault tolerance techniques and multi-version techniques along with Recovery blocks, N-version programming, N self-checking programming, consensus recovery blocks are major prevailing techniques.

CONSTRUCTION OF WEB ONTOLOGIES FROM RELATIONAL DATABASES


By:

          Abdul Sattar [MS-08-07]

Supervised By:

          Dr. Maruf Pasha

In this research work, we present a practical approach of automatic OWL ontology construction from relational database. The proposed approach will generate classes, subclasses, object properties, datatype properties, and cardinality restrictions from relational database schema automatically. First; we build domain concepts, and then learn concept relations, properties, and restrictions. Second, compare ontology created from normalized (3NF) and un-normalized (INF) Relational Databases.

SEMANTIC WEB ANALYSIS; DYNAMIC WEB CACHE MODELING


By:

          Muhammad Asif Javed [MS-08-17]

Supervised By:

          Dr. Maruf Pasha

Int his research, we aim to implement dynamic contents which is required during database interaction. The goal is to present HTML file to user instead of rendering pages every time from database. Results has proven that our proposal has reduced the response time, memory usage as it does not required database connection and holding results into Random Access Memory (RAM).

ROLE OF MULTI-AGENTS IN WEB SERVICES: A CASE STUDY


By:

          Adeeba Rashid Khan [MS-10-17]

Supervised By:

          Dr. Maruf Pasha

In this research, an architecture is proposed using the multiagent technology. Each agent is designed individually based on the FIPA standard agent platform, that includes the management system (MS), the agent directory (AD) which is the agent information repository in our architecture, and for the agent communication ACL syntax is used in designing their message communication. The ontologies are used to represent the domain knowledge. For ease the architecture is explained through a case study of medical field “Disease diagnostic system”. The case study problem represents completely the efficiency of multiagent technology.

AN APPROACH TO TEXTUAL EMAIL CLASSIFICATION USING SIMILARITY MEASURE: A COMPARATIVE STUDY


By:

          NasrUllah Khan [MS-10-18]

Supervised By:

          Israr Hanif

In this research work, we are suggesting the use of vector based document to query similarity measurement technique. Under the scope of Vector Space Model, a method for document to query similarity measurement in textual email classification is Cosine Similarity Measure – a string comparison method. Classifying an email to its best suitable destination by achieving higher classification accuracy is a fundamental goal of each method. A comparative study is performed among the previously used classification methods and our proposed approach. We have concluded from the results that the proposed approach is simple and effective due to its surprising precision and accuracy values as compared to other techniques previously used.

CASSANDRA MODEL FOR WEB APPLICATIONS


By:

          Qamar Hafeez [MS-10-05]

Supervised By:

          Dr. Qaiser

This research will help us to choose a better Database Management System (DBMS) when creating an enterprise level application in future. We have defined five different types of workloads in order to compare the performance of Apache Cassandra and MongoDB. In each workload, we used both databases and finally, in the end, conclude which of the two databases gives better performance results in which type of applications.

INFORMATION RETRIEVAL USING ONTOLOGY


By:

          Muhammad Ahsan Raza [MS-08-03]

Supervised By:

          Dr. Maruf Pasha

To achieve the context aware information retrieval by reformulating user queries, we have formulated an architecture followed by its working prototype. We have used OWL as basis of Computer Science ontology and Jena in retrieving context from ontology. The architecture is implemented using Java framework. The performance of the system is improved considerably using contextual information retrieved form domain specific ontology than those of conventional keyword based information retrieval.

COMPARATIVE ANALYSIS OF SERVICE ORCHESTRATION LANGUAGE: E-PLANNING ADMISSION SYSTEM OF PAKISTANI UNIVERSITY (OPUA)


By:

          Sehrish Hina [MS-10-16]

Supervised By:

          Dr. Maruf Pasha

In this research, an approach of design to compose different university services into a single platform (orchestrated web service) is proposed. The principle objective to collaborate the service is to use admission services offered by the institute. User can plan to get admission on program or course in desired university by having all details of admission policies and offered courses via orchestrated web service. User profile is used to help the proposed system for giving suggestions according to interest and preferences. The proposed system has service provider (universities services), service brokers (registered repository of services), user interface embedded with web application, orchestration service (multiple service workflow). WSM (web service mediator) integrate several different university information sources semantically annotated. SM (semantic mediator) collects the information from new registered and UDDI service repository, and send to the WSM. OWL-S profile’s describes each service capability and it has an ability to discover supporting service. OWL-S API is used to access to write, read and execute OWL-S orchestrated service description.


INTRA-DOMAIN ROUTING PROTOCOLS’ PERFORMANCE UNDER VARIOUS CONSTRAINTS


By:

          Ahmad Karim [MS-08-08]

Supervised By:

          Muzaffar Hameed

We have designed two network models that are configured with RIP, EIGRP, IGRP and OSPF routing protocols, to evaluate the performance of intra-domain routing protocols,. First scenario is built to measure performance in terms of some quantitative metrics such as network traffic sent and received, IP processing delay, packet loss ratio, CPU utilization, point-to-point throughput and point-to-point queuing delay for RIP, EIGRP, IGRP and OSPF protocols, through the simulated network models. The second scenario focused on the performance evaluation of RIP, IGRP, EIGRP and OSPF for real-time application demands such as voice, video streaming and data traffic such as FTP, HTTP, Database and Email applications under FIFO queuing technique. The parameters considered in this scenario are, response time of each application, queuing delay, queuing delay variation, response time, CPU utilization, TCP delay and TCP segment delay. The evaluation results show that EIGRP and OSPF performs better for both real-time and data applications than other routing protocols for the parameters considered.

DETECTION OF GRAY-HOLE ATTACKS IN MANETS


By:

          Aiza Shabbir [MS-08-10]

Supervised By:

          Dr. Qaiser Rasool

The main concern of this research is to develop a mechanism that provides an extended routing protocol and also points out the reasons for packet loss. Packet loss may be due to any specific attack, or due to congestion or lossy channel. Next when a node is detected as malicious then that particular node is identified and isolated from the network through proper techniques. Target is to stop the malicious node activity before it flow through out the whole network. The mechanism stops flows to the specific node and remaining flows are maintained unaffected and efficient. In order to cover up the loss of reliable packets, QOS is also increased.

SPAM DETECTION IN SOCIAL NETWORKING SITES


By:

          Nasira Perveen [MS-10-20]

Supervised By:

          Dr. Qaiser Rasool

In this research work, we have proposed some sentimental based features for spam detection. We collected 29K tweets of 29 most trending topics of 2012. We use classifiers like BayesNet, Naive Bayes, Random Forest, Support Vector Machine (SVM) and J48. Naive Bayes accuracy results improved to 11.18% with our proposed features in combination with baseline features. Random Forest accuracy is improved to 0.48%. J48 accuracy is improved to 0.47%. SVM accuracy is also improved to 0.14%. The experimental result shows that our proposed features can provide better classification accuracy results in combination with baseline content & user features by using Naive Bayes, Random Forest, J48 and SVM.

 

Please email your queries to the Chairman, Department of Computer Science, BZU.
or to this email address