Department of Computer Science and Engineering (CSE)
http://dspace.uiu.ac.bd/handle/52243/10
2024-03-28T22:42:28ZAnomaly Detection in Blockchain Transactions using Machine Learning with Explainability Analysis
http://dspace.uiu.ac.bd/handle/52243/2958
Anomaly Detection in Blockchain Transactions using Machine Learning with Explainability Analysis
Hasan, Mohammad
In the era of growing cryptocurrency adoption, Blockchain has emerged as a leading player in the digital payment landscape. However, this widespread popularity also brings forth an array of security challenges, including the need to safeguard against malicious activities. One of the paramount challenges in this regard is the detection of anomalous transactions within the
realm of Bitcoin data, a task that significantly influences the trust and security of digital payments. Yet, it’s a formidable challenge given the relatively low occurrence of anomalous Bitcoin transactions. Although several studies have been conducted in the field, a limitation persists: the lack of explanations for the model’s predictions. This study aims to address this limitation by combining eXplainable Artificial Intelligence (XAI) techniques and anomaly rules with tree-based ensemble classifiers. While deep learning
techniques have demonstrated their prowess in anomaly detection, there remains a scarcity of studies exploring their potential, particularly in the context of Bitcoin. This study also aims to fill that gap, focusing on our 1D Convolutional Neural Network (CNN) model. To understand how our model works and explain its decisions, we use the Shapley Additive exPlanation (SHAP) method, which measures each feature’s impact. We also deal with data imbalance by exploring various methods to balance anomalous
and non-anomalous Bitcoin transaction data. Additionally, we have introduced an under-sampling algorithm named XGBCLUS, designed to balance anomalous and non-anomalous transaction data. This algorithm is compared against other commonly used under-sampling and over-sampling techniques. Our experimental results demonstrate that: (i) XGBCLUS enhances TPR and ROC-AUC scores compared to state-of-the-art undersampling and over-sampling techniques, and (ii) our proposed ensemble classifiers outperform traditional single tree-based machine learning classifiers in terms of accuracy, TPR, and FPR scores, and (iii) our proposed 1D CNN model attains elevated accuracy with a concurrent reduction in the False Positive Rate (FPR).
2024-03-23T00:00:00ZHalal Food Identification from Product Ingredients using Machine Learning
http://dspace.uiu.ac.bd/handle/52243/2852
Halal Food Identification from Product Ingredients using Machine Learning
Tarannum, Sabrina
Halal food plays a critical role in the Islamic faith, as it represents food that is considered
lawful according to Islamic law. Muslims are encouraged to eat only Halal foods to ensure
that it aligns to their religious beliefs. However, locating and verifying Halal-certified foods
can be challenging, especially for Muslim travelers unfamiliar with the local food market.
Muslims ensure Halal foods that ingredients are prepared in accordance with Islamic
Shariah law. Indicators like the Halal emblem have been used to help Muslims identify
Halal food. Unfortunately, many packaged items are not Halal-certified. To address this
issue, this study presents a method for detecting Halal items using deep learning and
machine learning techniques. The purpose is to determine if an unknown product is Halal
(legal) or Haram (Illegal) based on its ingredients. The suggested system examines
packaged food product images and identifies the ingredients using the Yolo v5 algorithm.
The text on the images of the ingredients is then recognized using optical character
recognition (OCR). Various machine learning algorithms, artificial neural networks, and
fuzzy interference rule are applied to determine the status of the food. The final outcome is
to categorize Halal and Haram food products accurately. This approach has the potential to
assist Muslim consumers in identifying Halal-certified products quickly and efficiently,
particularly when traveling to new locations or encountering unfamiliar products. Using
intelligent technology; this study presents a new and innovative technique for detecting
Halal food. The result shows that the suggested approach is effective and it might be a
useful tool for Muslim consumers in ensuring that the things they buy are compatible with
their religious views
2023-09-09T00:00:00ZMemory Forensics for Analyzing Malicious Activities
http://dspace.uiu.ac.bd/handle/52243/2835
Memory Forensics for Analyzing Malicious Activities
Prottoy, Rafid Asrar
With the change of era, the growing dependency on the computer and Internet is needless to say in a word. Memory is a very important part of a computer that holds the necessary data that the processor uses. As the CPU's running process data is stored in the memory, capturing and preserving the memory information are very important to detect malicious activities. If the memory is volatile like in RAM, data can be easily lost by overwriting or power failure. So, creating the memory dump from the volatile and secondary memory is invaluable for memory forensics and identifying different malicious activities for forensic investigation. Memory dump information can be used forensically to detect malicious activities within the suspected device. Nowadays, Internet usage is increasing tremendously, so people face many attacks like malware originated from the Internet. The attacker uses the victim's machine to execute their plan anonymously. During the investigation, there will be voluminous amount of information to investigate. As malicious processes are smart enough to hide, finding the malicious processes are not that trivial. Investigators must relate the incident data from the memory dump information to identify the malicious activities. There are many challenges in creating the memory dump from the heterogeneous types of devices and investigating the collected memory dump if investigators do not use the right methods and tools, which will enable to create the memory dump mellifluously and to identify different malicious activities in a short time. Using the right tools and frameworks at the right time, the effectiveness of the investigation can be much better and faster. In traditional processes, there is no structural way to find malicious activity. So, in this project, we have proposed a method for investigating malicious activities in a more structured and efficient way from the captured memory dump and identifying malicious activities from a suspected machine.
2023-07-31T00:00:00ZPCA-ANN: Feature Selection based Hybrid Intrusion Detection System in Software Defined Network
http://dspace.uiu.ac.bd/handle/52243/2829
PCA-ANN: Feature Selection based Hybrid Intrusion Detection System in Software Defined Network
Nawshin, Sabila
The increasing complexity of modern networks and the rise of sophisticated cyber attacks has made the development of effective Intrusion Detection Systems (IDS) a critical need. Software De fined Networking (SDN) technology provides us with a programmable central controller, providing a central view of the whole network as opposed to the existing internet structure where each of the routers only has information about it's surrounding routers, which results in the systems and algorithms developed in it to operate
in an distributed setting. The centralized view provided by SDN makes it an attractive platform for IDS deployment. The networks under SDN is, however, more vulnerable to malicious activities or attacks than the traditional network topology due to the same centralised nature. The recently published "inSDN" dataset was prepared specifically for intrusion detection in SDN. In this study, we have used this dataset to introduce a novel Intrusion Detection System (IDS) model that integrates Principal Component Analysis (PCA) - a feature selection methodology commonly used in traditional Machine Learning (ML) to extract the principal features from large datasets and reduce dimensionality - and Artificial Neural Networks (ANN) to classify network tra c based on the extracted features. The
model achieved an accuracy of 99.95% for multi-class classifi cation. The results show that the proposed model outperforms the current state-of-the-art techniques in a much simpler settings and reduces the need for complex models that require extensive computation in the "inSDN" attack dataset.
2023-07-01T00:00:00Z