Distributed Denial-of-Service (DDoS) attacks continue to pose a significant threat to digital infrastructures, often resulting in degraded service availability and financial losses. Traditional detection systems, which depend on static rule sets, struggle to adapt to evolving traffic patterns, leading to increased false positives and undetected attacks. This paper presents a real-time, machine learning-based framework for DDoS detection and mitigation. The framework incorporates supervised learning algorithms, including Random Forest, XGBoost, and Multi-Layer Perceptron (MLP), trained on the CIC-DDoS2019 dataset using carefully selected network traffic features to enhance detection accuracy. The system architecture integrates Scapy for traffic capture, Apache Kafka for message queuing, and Flask with Plotly for dynamic monitoring. Evaluation results demonstrate superior performance compared to legacy methods across precision, recall, F1-score, false positive rate (FPR), and false negative rate (FNR). Additionally, adaptive models such as Passive-Aggressive and Stochastic Gradient Descent (SGD) enhance robustness against evolving attack vectors. The proposed solution delivers an effective and scalable real-time defense mechanism suitable for banking, cloud, and enterprise systems. However, the system’s performance remains influenced by the characteristics of the training dataset and may introduce computational overhead during high-throughput traffic analysis. Future work will explore improved computational efficiency and responsiveness to rare or emerging DDoS patterns.

Social media usage has increased due to the rate at which technologies are emerging and it is less likely to detect false news/information manually as it aims to capture the human mind. The spread of false news can cause havoc; therefore, detection of false news becomes paramount where almost everyone has access to social media. Our proposed system optimizes the false news detection process. The system combines advantages of two textual feature extraction methods and two machine learning algorithms for text classification. Basic pre-processing methods were employed. Feature extraction was carried out using Term Frequency-Inverse Document Frequency with Word2Vector. K-Nearest Neighbour (KNN) and Naïve Bayes (NB) algorithms are combined to give KNN Bayesian. The most available systems made use of a single feature extraction method but in our system, two feature extraction methods are combined. The evaluation metrics used were accuracy, precision, recall, f1score and KNN Bayesian performed better than KNN. To further evaluate our model, the Area under the Curve-Receiver Operator Characteristics (AUC-ROC) revealed that AUC of KNN Bayesian ROC curve is higher than that of KNN.