Muhammad Junaid Iqbal and Jordi Serra-Ruiz, Universitat Oberta de Catalunya (UOC), Spain

Ransomware has emerged as a critical and rapidly evolving cybersecurity threat, significantly impacting sectors such as healthcare, finance, and government infrastructures. This paper presents a comprehensive survey of contemporary ransomware detection techniques, focusing on machine learning (ML) and deep learning (DL) methodologies, which have shown promise in adapting to the rapidly changing landscape of ransomware attacks. The survey includes a detailed comparative analysis of static, dynamic, and hybrid detection models, highlighting their respective advantages and limitations. The key findings from the survey show that ML and DL-based methods have a better detection capabilities but still having challenges such as large and diverse datasets, the computational cost of advanced techniques, and model adaptability across various platforms still exist. We also delve into some up-and-coming trends, like quantum computing and federated learning, both of which have the potential to overcome present limitations in computation efficiency and privacy concerns, respectively. It also points to the increasing attention being paid to adversarial defenses, which seek to make models more robust against complex evasion attempts.

Ransomware; Machine learning; Deep Learning; Cybersecurity; Hybrid Detection; Adversarial Defense