A Comparative Analysis of Supply Chain Predictive AI in the Defense Industrial Base
Open Access DepositedRecent studies such as those by Osamor et al. (2025) confirm that supply chain security is a critical indicator of cybersecurity resilience for organizations that have successfully mitigated security risks. Cyber resilience has become especially relevant in the United States defense sector, where supply chain security has become a focal point in recent legislation such as Executive Order 14017 (The United States Government, 2021). Global imports and exports dropped to the levels of the 2008 Financial Crisis due to supply chain disruptions during the COVID-19 pandemic (Giovanni et al., 2022). According to Statista's Cybercrime Statistics, supply chain attacks have steadily increased from 69 in 2020 to 242 in 2023 (Identity Theft Resource Center, 2024). The impact of these disruptions has been significant for the United States Defense sector (The United States Government, 2021). In response, this research aims to compare machine learning models to detect malware in the cyber supply chain and enhance the security of the Defense Industrial Base (DIB). Additionally, it identifies critical cyber supply chain features to inform DIB technology purchases to aid in the federal acquisition process. Selecting the appropriate machine learning model for effective detection and prevention of cyber supply chain attacks can support federal procurement strategies within the Defense Industrial Base (DIB). Ultimately, the findings from this research will provide concrete recommendations to improve strategic decision-making and enhance cybersecurity in federal acquisitions.This research integrates the Cyber Threat Intelligence (CTI) identified in the literature review to provide context for interpreting the classification results after applying the chosen models to the Classification of Malware (CLAMP) dataset. The information in the dataset comes from the header field values of Portable Executable (PE) files provided by Microsoft which is a key member of the DIB’s cyber supply chain. Kumar (2020) used the dataset to generate derived features which serve as an input to this research analysis. The most critical features identified are used to mitigate the risks in the cyber supply chain of DIB to improve its overall security and resilience. This research applies supervised feature selection ANOVA, Kruskal-Wallis, Chi-Squared, and Correlation-based Feature Selection (CFS), to determine which features are most influential in malware classification. The study also evaluates the performance of eight types of supervised machine learning models in predicting malware threats, which are tree-based, kernel-based, instance-based, linear, logistic, subspace-based, gradient boosting, and neural network, along with two unsupervised methods, K-Means, and hierarchical clustering. To reduce the risk of overfitting, this study evaluates the performance of seventeen specific machine learning models from the eight groups mentioned above using 10-fold cross-validation. After comparing and evaluating the models using the following performance metrics
accuracy, precision, recall, speed, and F1-score, the Bagged Decision Tree model showed the highest accuracy. Moreover, the supervised models outperform the unsupervised models in identifying malware. The overall efficiency of supervised models proves that machine learning can be helpful for early threat detection and risk mitigation. While this research is specific to PE malware threats, the models can also contribute to developing broader DIB strategies to improve cybersecurity defenses and support increased resilience. This proactive approach enhances individual organizations' security postures and contributes to the broader protection of national defense infrastructure.
- 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.