Theses

Now showing 1 - 3 of 3

Embargo
Mitre attack framework adoption as a siem rule base using machine learning approach
(2021) Weeraman, P.W.R.S.
Digital transformation is the standard business strategy approach in most Organizations. Every person is looking for digital solutions to aid their routine works. Every Organization looking possibility move to physical office concept for virtual office concept. Even homemakers and bargain hunters also expect to move online shopping with doorstep delivery solutions with this COVID-19 pandemic. Every business needs to adopt IT functions for their business process to ensure business stability or increase their revenue. Most large-scale enterprises have a dedicated IT strategy approach to align with their business strategy. They follow best IT security practices such as SIEM, security operation centers (SOC), annual IT compliance review, IT audit and best security devices in the market. However, most of the business do IT system adoption without a preplanned process. They do not follow any best it practices in term of IT security. Further, they do not have a proper IT strategy that aligns with business objectives. Most small and medium scale business with minimum IT infrastructures and IT operations. The absence of a proper IT security approach in the business may introduce new IT risk to their information and business. This Research makes experimental approach to adopt cyber threat intelligence to SIEM detection base using adversary tactic, technique, procedure (TTP) and machine learning (ML) instead of signature-based detection methods. TTP change is relatively more challenging than IP address or file hash change. This research concern uses TTP-based Security information and event management systems (SIEM) solution using open-source software and MITRE ATT&CK community framework. Further, this Research aims to reduce operating expenses and capital expenses using a community-based framework and opensource software.
Embargo
Decision Support System for Overcoming the Challenges in Vocational Education in Sri Lanka
(2021) Lakshani, J. K. A. M.
The vocational education is undergoing continuous changes. In the past, high youth unemployment has taken place due to unfamiliarity with vocational education. Researchers and policy makers are paying attention to the vocational education because of the hidden importance of the vocational education. In Sri Lanka, there is a vocational education system as the 13 years mandatory education system. The project is going to discover the challenges of the vocational education and give some solution to enhance the effectiveness of vocational education using the sample scenario of the professional entry. There are several issues in vocational education system. Among them, the major challenge is the lower rate of successfully completed students than commencing students. The main objective of this research is to develop a Data-driven decision support system to mitigate the students’ dropouts from vocational education using deep learning model with higher level of accuracy rate than previous systems. Accurate data collection helps to maintain the integrity of the research in any field. The project has collected real data set from the students and teachers in selected government schools in Sri Lanka. Data has collected mainly in three categories as demographic factors, academic performance and candidate interest. Collected data has analyzed according to the data analysis techniques. Decision support system has used machine learning model to predict the suitable vocational education pathways to the students. The model has used deep neural network (DNN) with PyTorch library. After training the model, the model has predicted the accuracy level as 96.06%.
Embargo
Unsupervised Sinhala Cyberbullying Categorization
(2021) Chandrasena, B.G.M
The objective of unsupervised machine learning is to categorize the social media comments into a given number of pre-learned categories. The earlier studies of this domain have used many the dataset for supervised learning & introduced a large number of techniques, methodologies. A major challenge there was training labels. Although words with training comments are easy to find, separating them manually is not an easy task. Through this research, we hope to find a solution to this using unsupervised machine learning techniques. the proposed technique divides the comments into words and removed special characters, emojis, and links from the comments & categorized each comment using a keyword list of each category and similarity findings. And then this was used to categorize comments for training. The implemented method shows the same performance, by Comparison with other supervised machine learning techniques for cyberbullying. Therefore, this mechanism can be used in any other places where low-cost cyberbullying identification is needed. This also can be used to create train comments.

Theses

Browse

Filters

Advanced Search

Filter by

Settings

Sort By

Results per page

Search Results