A Machine Learning Approach to Evaluate Ransomware and Data Breach in Higher Education Institutions
Open Access DepositedAbstractThe academic sector, marked by its vast and diverse sensitive data, has emerged as a significant target for cybercriminals in the digital era. This trend is underscored by a 44% escalation in cyber-attacks on academic institutions from 2021 to 2022, culminating in a distressing weekly average of 2,297 incidents in 2022. The swift shift to digital learning, catalyzed by the COVID-19 pandemic, has amplified this concern. In response to the escalating threat of digital attacks, this research applies machine learning methodologies to identify and quantify the primary factors associated with two prevalent types of cyber threats: ransomware and data breaches. The study consolidated demographic data and historical records of reported attacks from 2,297 U.S.-based academic institutions, leading to a comprehensive analysis. Given the disparity in the dataset—with 636 confirmed incidents contrasted against the total number of institutions—advanced data analytics and machine learning models were employed. Four distinct machine learning models—Logistic Regression, Voting Ensemble, Random Forest, and Light GBM—were evaluated, validated, and trained. Their efficacious performance under such conditions is assessed through rigorous cross-validation and evaluation using several metrics, including weighted accuracy, precision, recall, and F1-score. Notably, the Voting Ensemble model outperforms the others, achieving an outstanding weighted accuracy score of 0.92290. Insights from the robust Voting Ensemble model reveal key factors distinguishing institutions with a history of cyber-attacks from those without any recorded incidents. The size of an institution emerges as the most influential factor, followed by the presence of a medical school, the number of degrees offered, research status, and academic ranking. By employing advanced data analytics, this study delivers an in-depth analysis of characteristics that increase the risk of specific types of cyber-attacks in academic institutions, thereby laying a robust foundation for further research. Future research directions include augmenting the dataset with underrepresented classes to enhance classification accuracy and broadening the scope to include K-12 schools and similar entities in different sectors. Such expansion would provide more comprehensive insights and uncover additional key factors associated with ransomware and data breach attacks. The findings of this research significantly contribute to the existing body of knowledge and have crucial implications for the security of academic institutions in the digital age.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.
| Thumbnail | Title | Date Uploaded | Visibility | Actions |
|---|---|---|---|---|
|
|
Thompson_gwu_0075A_16540.pdf | 2024-10-02 | Open Access |
|