Design and Implementation of a Fuzzy Intelligent Instant Selection Algorithm for Intrusion Detection Systems

Corrected Title: Design Implementation of a Fuzzy Intelligent Instance Selection Algorithm for Intrusion Detection Systems One member replacement: Ifra Arshad (CNIC 3620153207228    Cell Number 0349-6706287    email

Project Title

Design and Implementation of a Fuzzy Intelligent Instant Selection Algorithm for Intrusion Detection Systems

Project Area of Specialization

Cyber Security

Project Summary

Corrected Title:

Design Implementation of a Fuzzy Intelligent Instance Selection Algorithm for Intrusion Detection Systems

One member replacement:

Ifra Arshad (CNIC 3620153207228    Cell Number 0349-6706287    email ifraarshad611@gmail.com)

with

Usama Ejaz (CNIC 3220205149705    Cell Number 0304-1404022     email  usamaejaz418@gmail.com)

Building a high quality classi?er is one of the key problems in the ?eld of machine learning (ML) and pattern recognition. Many ML algorithms have suffered from high computational power in the presence of large scale data sets. Inspired by the work “Rana Aamir Raza Ashfaq, Yu-lin He, De-gang Chen, Toward an ef?cient fuzziness based instance selection methodology for intrusion detection system, Int. J. Mach. Learn. & Cyber. (2017) 8:1767–1776”, the present project is to implement a fuzziness based instance selection technique for the large data sets to increase the ef?ciency of supervised learning algorithms by improving the shortcomings of designing an effective intrusion detection system (IDS). The methodology is dependent on a new kind of single layer feed forward neural network (SLFN), called random weight neural network (RWNN). At the ?rst stage, a membership vector corresponding to every training instance is to be obtained by using random weight neural network (RWNN) for computing the fuzziness. Secondly, the training instances (along with their fuzziness values) according to the actual class labels are to be grouped separately. After this, the instances having low fuzziness values in each group are to be extracted, which would be used to build a reduced data set. The instances outputted by the said methodology are to be used as an input for machine learning (ML) classi?ers, which would result in reducing the learning time and also increasing the learning capability. The methodology exhibits that the reduced data set can easily learn the boundaries between class labels. The most obvious ?nding from this study would be a considerable increase in the accuracy rate with unseen examples when compared with other instance selection method, i.e., IB2. The methodology provides a better generalization and fast learning capability. The usefulness of the methodology is to be demonstrated with experiments on some well known ID data sets.

Project Objectives

The methodology to be implemented is dependent on a learning algorithm that is used for constructing the intrusion detection system model. Inspired by the work “Rana Aamir Raza Ashfaq, Yu-lin He, De-gang Chen, Toward an ef?cient fuzziness based instance selection methodology for intrusion detection system, Int. J. Mach. Learn. & Cyber. (2017) 8:1767–1776”, a small subset is obtained which is used as a representation of whole training data set. For this purpose, fuzziness is considered as an important criterion for the selection of instances. In the methodology, the plan is to consider fuzziness quantity as an important criterion for the selection of training instances. Therefore, the main objective is to propose a methodology for optimizing the training data set in such a way that, new training data can easier to learn the boundaries between class labels, i.e., normal and anomaly, in intrusion detection systems. The random weight neural network (RWNN) is to be used as a base classi?er, to obtain the membership vector corresponding to every training instance. After computing the fuzziness of every membership vector, the instances are grouped according to the actual class labels along with their fuzziness values. Only extract those training samples from each group that are having low fuzziness value, and then, train extracted samples with RWNN and other AI algorithms to evaluate their performance.

Project Implementation Method

At the ?rst stage, a membership vector corresponding to every training instance is to be obtained by using random weight neural network (RWNN) for computing the fuzziness. Secondly, the training instances (along with their fuzziness values) according to the actual class labels are to be grouped separately. After this, the instances having low fuzziness values in each group are to be extracted, which would be used to build a reduced data set. The instances outputted by the said methodology are to be used as an input for machine learning (ML) classi?ers, which would result in reducing the learning time and also increasing the learning capability.

Benefits of the Project

The most obvious ?nding from this study is targeted be a considerable increase in the accuracy rate with unseen examples when compared with another instance selection method, i.e., IB2.

The methodology to be implemented would also provide a better generalization and fast machine learning capability.

Technical Details of Final Deliverable

The final deliverable of the project is Research Report and Journal Publication.

Final Deliverable of the Project

Software System

Type of Industry

IT , Security

Technologies

Artificial Intelligence(AI), Big Data

Sustainable Development Goals

Industry, Innovation and Infrastructure, Sustainable Cities and Communities

Required Resources

If you need this project, please contact me on contact@adikhanofficial.com