Copyright (c) 2025 Working Paper ECBTI

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
Artificial intelligence and information security: towards a predictive model of digital protection. A systematic literature review
Artificial intelligence (AI) now occupies a central place in digital protection strategies, yet the accelerated growth of its literature makes it difficult to obtain an integrated picture of its technical, operational, organizational, and ethical components. This study addresses the absence of a conceptual framework explaining how the convergence of machine learning, deep learning, algorithmic explainability, and technology governance enables a transition from reactive to predictive cybersecurity, through a systematic literature review adapted from the PRISMA approach. The study examined 340 records identified in Scopus, IEEE Xplore, ScienceDirect, Web of Science, Dialnet, and institutional repositories between 3 and 28 February 2025; 37 studies were included in the final synthesis and 20 are directly cited. The methodology combined thematic content analysis, a data extraction matrix, and explicit quality criteria, together with a co-occurrence analysis of 15 key terms that revealed three conceptual clusters: technical, operational, and ethical-regulatory. The results show that predictive and hybrid models substantially improve the capacity to anticipate threats, although their effectiveness depends on training data quality and human oversight, and reveal a widening gap between technical sophistication and institutional readiness. Based on these findings, the study proposes the Integrated Framework of AI-Based Predictive Cybersecurity (MICPIA), structured into four interrelated dimensions -technological, operational, organizational, and ethical- and grounded in five theoretical propositions. The study concludes that the advancement of AI-driven cybersecurity depends as much on algorithmic progress as on organizations' capacity to govern it under explicit ethical criteria, and that the original contribution of this work lies in integrating, within a single framework, technical evidence and normative requirements the literature usually treats separately