TITLE:
Artificial Intelligence Systems Cybersecurity Ensuring: Analysis of Vulnerabilities, Attacks, and Countermeasures
AUTHORS:
Conrad Onesime Oboulhas Tsahat, Ngoulou-A-Ndzieli
KEYWORDS:
Artificial Intelligence, Cybersecurity, Vulnerability, Attacks, Machine Learning
JOURNAL NAME:
Journal of Information Security,
Vol.17 No.1,
December
22,
2025
ABSTRACT: The rapid adoption of Artificial Intelligence (AI) systems in critical sectors of society has given rise to new cybersecurity challenges. Unlike traditional software systems, AI systems have unique characteristics such as data dependence, model complexity, and adaptive behavior, which create new types of vulnerabilities and attack vectors. Through such attacks, intruders can manipulate these systems to change their behavior to achieve their goals. According to expert data, only 25% of modern artificial intelligence applications are properly protected. Given these technologies specifics their security covers a wide range of tasks, including the data protection, algorithmic models and application scenarios. This review article provides a comprehensive analysis of the current state of AI cybersecurity, systematizing the vulnerabilities inherent in AI, classifying the main types of attacks at different stages of the AI lifecycle, and describing adequate countermeasures. This paper proposes a comprehensive taxonomy of threats and defenses, covering aspects from data collection to model deployment and operation. The goal of the paper is to provide a deep understanding of the complex AI threat landscape and guide researchers and practitioners in the development and implementation of robust and secure AI systems. Finally, current research gaps are identified and future directions are outlined to ensure the sustainability of AI in a dynamically changing digital environment.