International Journal for Electronic Crime Investigation

Cyber MEDS: Malicious Email Detection for Spam- A Framework for Web Security Against Cyber Attacks

2025-10-25T11:37:12+00:00

Email is still one of the main ways cybercriminals attack, especially through spam and phishing messages. These unwanted emails are not just an annoyance, it can lead to serious risks such as stealing sensitive data, financial fraud, spreading harmful software, etc. This creates a constant security challenge, for both individuals and organizations. In this study, design a practical and efficient framework for classification of spam emails using multiple machine learning techniques. The study compared several algorithms, including Random Forest, Gaussian Naive Bayes, Multi-Layer Perceptron, Gradient Boosting, and K-Nearest Neighbors, on the well-known public Spambase dataset. Apply Min-Max scaling to make all features fall in the same range, which helps the learning process and improves prediction quality, before model training. The experimental results show that the Random Forest model gives the best overall performance, achieving 95.11% accuracy, 95.89% precision, 91.34% recall, and 93.56% F1-score. These results show that even lightweight, carefully tuned models can detect harmful emails with high reliability, providing an early layer of defense in email security. Study also adds to the growing research on building scalable, dependable solutions that can adapt to the constantly changing nature of Cyber threats.

Big Data Analytics and Machine Learning Techniques for Real-Time Credit Card Fraud Detection

2025-10-25T11:30:24+00:00

Big Data is commonly characterized by the 4 V's: Volume, Variety, Velocity, and Veracity. In today’s digital age, data is generated in terabytes and petabytes, far exceeding the storage capabilities of a single machine. With data constantly circulating across cloud platforms, the risk of leakage and fraud has increased significantly, with credit card fraud being one of the most pressing global concerns. As numerous shopping platforms and businesses operate around us, each domain generates vast amounts of data, often reaching into yottabytes. Manually handling, analyzing, or detecting anomalies in such large-scale data is extremely challenging. However, with the advancement of computing and emerging technologies, detecting fraud has become much more efficient and scalable. This study examines the application of big data in analyzing credit card consumer behavior, specifically in the context of online transactions, password creation, age, income, and other relevant factors. The focus is on identifying anomalies in these data points to detect potentially fraudulent activities quantitative approach is employed to identify statistical patterns, and the performance of seven different machine learning algorithms, such as Logistic Regression, K-Nearest Neighbors (KNN), and XGBoost, is evaluated for their effectiveness. As technology advances, factors such as age and increasing reliance on online transactions, e-commerce, and digital banking contribute to rising vulnerabilities, making fraud detection more critical than ever. In the Real-time credit card fraud detection using big data, different algorithms are discussed and implemented so XGBOOST gives better results with 99% accuracy as another ML Algorithm. The impact of compliance on sophisticated data-based security systems will be examined in a later study, which can make use of historical fraud typologies and trends to comprehend potential shifts over time.

Securing 5G Network Infrastructure Against DDoS Threats Using ML-Based Anomaly Detection

2025-10-25T11:30:21+00:00

Today, millions of people and devices use the Internet to carry out daily activities, but the growing reliance on the Internet comes with major security concerns. Older security systems and traditional detection techniques are out of date because attackers continue to find new and smarter ways of penetrating networks. They are just not precise enough to stay in the race. This research discusses how that gap can be filled by machine learning (ML). Although in cybersecurity, ML has demonstrated potential, accuracy remains reliant on the selection of the appropriate models and the concentration on the most important parts of the data. Although ML has already shown its potential, our work aims at refining the approach to increase detection accuracy. The most promising among the techniques tested was the Random Forest (RF) algorithm, which had an impressive accuracy rate of 99.84%. This clearly indicates that our proposed system is far better than the previous methods, showing its capability to detect malicious activities.

Efficient Blind Multi Receiver Signcryption of Secure Multicast in IoT and Beyond

2025-10-25T11:39:29+00:00

This research introduced Blind Multi-Receiver Signcryption (BMRSC) scheme which is designed upon Elliptic Curve Cryptography (ECC) to improve security and privacy in networks with limited computation powers. The protocol also integrates Blind signature and signcryption protocol to enable one-to-many secure communication that in particular, is applicable to electronic voting and electronic currency as well as the Internet of Things (IoT) networks.. The scheme has lightweight ECC operations and therefore has small computational and communication overheads, which are the major resources for implementing a scheme on mobile and embedded devices. The scheme not only ensures confidentiality, authenticity and anonymity of the sender, but it also supports forward secrecy and unlinkability properties, which are not provided in other designs. Security analysis is employed to ensure resilience to vulnerabilities to critical threats such as forgery and key exposure attacks, and comparative analysis demonstrates that the proposed solution is more efficient than state-of-the-art blind signcryption protocols..