Malicious Uniform Resource Locators (URLs) pose a significant cybersecurity threat by carrying out attacks such as phishing and malware propagation. Conventional malicious URL detection methods, relying on blacklists and heuristics, often struggle to identify new and obfuscated malicious URLs. To address this challenge, machine learning and deep learning have been leveraged to enhance detection capabilities, albeit relying heavily on large and frequently updated datasets. Furthermore, the efficacy of these methods is intrinsically tied to the quality of the training data, a requirement that becomes increasingly challenging to fulfill in real-world scenarios due to constraints such as data scarcity, privacy concerns, and the dynamic nature of evolving cyber threats. In this study, we introduce an innovative framework for malicious URL detection based on predefined static feature classification by allocating priority coefficients and feature evaluation methods. Our feature classification encompasses 42 classes, including blacklist, lexical, host-based, and content-based features. To validate our framework, we collected a dataset of 5000 real-world URLs from prominent phishing and malware websites, namely URLhaus and PhishTank. We assessed our framework’s performance using three supervised machine learning methods: Support Vector Machine (SVM), Random Forest (RF), and Bayesian Network (BN). The results demonstrate that our framework outperforms these methods, achieving an impressive detection accuracy of 98.95% and a precision value of 98.60%. Furthermore, we conducted a benchmarking analysis against three comprehensive malicious URL detection methods (PDRCNN, the Li method, and URLNet), demonstrating that our proposed framework excels in terms of accuracy and precision. In conclusion, our novel malicious URL detection framework substantially enhances accuracy, significantly bolstering cybersecurity defences against emerging threats.

Its RESEARCH ARTICLE IN DOMAIN OF BLOCKCHAIN TECHNOLGY.

Interactive multi-objective optimisation allows a decision maker to steer a search toward preferred trade-offs while it runs, rather than selecting only from an already-computed Pareto front. Existing interactive methods typically represent preference through linear value functions or weights that are agnostic to risk, while methods that do account for risk either fix the risk attitude in advance or apply it only once the search has ended, providing no mechanism for estimating an individual decision maker’s attitude towards risk during the search. We address this gap with a framework that instead infers a decision maker’s attitude towards risk from choices made during the search itself and uses the inferred attitude to steer the remaining search directly. Risk attitude is represented via Constant Relative Risk Aversion (CRRA) utility, a standard economic model in which a single coefficient 𝜌 indexes whether a decision maker is risk-averse (𝜌 > 0) or risk-seeking (𝜌 < 0). We refer to the resulting framework as CREST (CRRA-based Risk Elicitation and Steering). At intervals during the search, the decision maker compares a pair of candidate strategies, each evaluated across fifty simulated scenarios; the resulting choice updates a continuous estimate of 𝜌 via softmax maximum-likelihood estimation, which treats choices as probabilistic rather than deterministic. Once this estimate is judged stable, it is used to steer the remaining generations of the underlying evolutionary search directly, favouring candidates according to the decision maker’s estimated utility rather than continuing to query the decision maker. CREST is applied to a Navy workforce planning problem and validated against a human decision maker and two independently trained large language models (LLMs), Llama 3.1 8B and Qwen3 8B, each instructed to adopt a risk-averse or risk-seeking persona. A stable risk attitude was identified within four rounds in all six decision-maker–persona conditions tested. Which LLM most closely resembles the human decision maker, however, depends on the persona being elicited, and an identical estimated attitude does not guarantee a matching workforce policy: under the risk-seeking persona, all three decision maker’s converge on the same extreme attitude, yet the resulting strategies diverge substantially between the human decision maker and both LLMs. An LLM’s suitability as a stand-in for a human decision maker must therefore be established empirically for the specific model and persona in use rather than assumed.