Phishing remains one of the most persistent cybersecurity threats, increasingly exploiting not only technical vulnerabilities but also human cognitive biases. Existing detection systems often rely on single-modality features and black-box models, which restrict both generalization and interpretability. This study presents an explainable multimodal framework that combines textual and technical cues, including message content, URL structure, and Principles of Persuasion, to capture both objective and subjective aspects of phishing. Several classifiers were evaluated using 10-fold stratified cross-validation, with Random Forest achieving the best balance between performance and transparency (ROC-AUC = 0.9840), supported by SHAP explanations that identify the most influential linguistic and structural features. Comparative analysis shows that the proposed framework outperforms unimodal baselines while preserving interpretability, enabling a clear rationale for classification outcomes. These results indicate that integrating multimodal representation with explainable learning strengthens phishing detection accuracy, improves user trust, and supports reliable deployment in real-world environments.

Phishing, a pervasive cybersecurity issue, involves fraudulent attempts to obtain sensitive information and to provoke unintentional money transfers or malware downloads, among others, by disguising as trustworthy entities in electronic communications. This paper presents an innovative approach to phishing detection in Spanish emails using patterns represented as rules. Through a comprehensive, still efficient analysis of emails, we identify interpretable recurring patterns and relevant phrases used in phishing attempts. These phrases and words often aim to persuade victims into revealing personal or financial information. These patterns are translated into a set of rules that are applied to evaluate incoming emails. Additionally, a proof-of-concept is carried out using a phishing data set of Spanish emails created for this study. Our method achieved promising results in identifying phishing attempts, providing an additional layer of security for email users. Moreover, this approach can be adapted to detect phishing in other languages, making it a potentially global solution to this persistent cybersecurity issue.

With the rising of Internet in early ’90s, many fraudulent activities have migrated from physical to digital: one of them is phishing. Phishing is a deceptive practice focused on exploiting the human factor, which is the most vulnerable aspect of any security process. In this scam, social engineering techniques are extensively utilized, specifically focusing on the principles of persuasion, to deceive individuals into disclosing sensitive information or engaging in malicious actions. This research explores the use of message subjectivity for detecting phishing attacks. It does so by assessing the impact of various data representations and classifiers on automatically identifying principles of persuasion. Furthermore, it investigates how these detected principles of persuasion can be leveraged for identifying phishing attacks. The experiments conducted revealed that there is no universal solution for data representation and classifier selection to effectively detect all principles of persuasion. Instead, a tailored combination of data representation and classifiers is required for detecting each principle. The Machine Learning models created automatically detect principles of persuasion with confidence levels ranging from 0.7306 to 0.8191 for AUC-ROC. Next, principles of persuasion detected are used for phishing detection. This study also emphasizes the need for user-friendly and comprehensible models. To validate the proposal presented, several families of classifiers were tested, but among all of them, tree-based models (and Random Forest in particular) stand out as preferred option. These models achieve similar level of effectiveness as alternative methods while offering improved clarity and user-friendliness, with an AUC-ROC of 0.859842.

Governmental adminstrative domains can potentially benefit from a wide variety of currently available big data analysis methods. The tax administration is such an area that requires massive data processing to identify hidden patterns and trends of possible tax evasion. The use of supervised methods can be effective in these cases, but the lack of available labeled data limits their practical application in real-world scenarios. An alternative is the use of unsupervised methods, which have potential benefits in certain cases. In this sense, unsupervised methods are considered to be feasible as a decision support tool in tax evasion risk management systems. This paper proposes an unsupervised approach to identify signs of tax evasion by detecting, possible, tax underreporting. The proposed strategy is evaluated on a data set associated with individual income tax statistics of the United States. The results achieved are considered to be useful in decision-making and preventive actions on cases reported as suspicious.

Persuasion is a human activity of influence. In marketing, persuasion can help customers find solutions to their problems, make informed choices, or convince someone to buy a useful (or useless) product or service. In computer crimes, persuasion can trick users into revealing sensitive information, or even performing actions that benefit attackers. Phishing is one of the most common and dangerous forms of persuasion-based attacks, as it exploits human vulnerabilities rather than technical ones. Therefore, an intelligent system capable of detecting and classifying persuasion attempts might be useful in protecting users. In this work, an approach that uses Machine Learning to analyze messages based on principles of persuasion and different data representations is presented. The aim of this research is to detect which data representation and which classification algorithm obtain the best results in detecting each principle of persuasion as a prior step to detecting phishing attacks. The results obtained indicate that among the combinations tested, there is one combination of data representation and classification algorithm that performs best. The related classification models obtained can detect the principles of persuasion at a rate that varies between 0.78 and 0.86 of AUC-ROC.