| Autor: |
Taha, Nameer Nail, Abdullah, Nada Abdulzahra |
| Předmět: |
|
| Zdroj: |
International Journal of Intelligent Engineering & Systems; 2024, Vol. 17 Issue 6, p663-675, 13p |
| Abstrakt: |
Attacks using persistent ransomware can be extremely harmful to both individuals and businesses. Even with the availability of powerful post-infection remediation techniques, ransomware offenders are now able to outwit conventional security defenses by using sophisticated pre-attack techniques, which makes early-stage detection critical. The detection of ransomware in its early phases is necessary to prevent additional damage from occurring. This paper presents a new framework for detecting ransomware. The approach utilizes a specialized pipeline for feature engineering and a neural network model for classifying multiple classes. The main innovation is the utilization of API call features in dynamic analysis. These features are designed to detect specific behavioral patterns that are exclusive to various types of ransomware. This approach enables the model to discern between benign and malicious actions with a high level of accuracy. The model under consideration was trained and evaluated using a dataset of 5203 samples from 13 different ransomware families. It achieved an accuracy rate of 92%. To assess the model's capacity to apply to different situations, it was tested on a more complex dataset containing five more types of ransomware. The model attained an accuracy rate of 98.22% in this evaluation. This framework surpasses current methods, especially when dealing with intricate, practical datasets. It showcases a strong ability to detect attacks at an early stage and offers a proactive defense against ever-changing cyber risks. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Complementary Index |
| Externí odkaz: |
|
