MICO ARTEFACT: DEVELOPMENT OF COMPONENTS FOR THE MACHINE INTELLIGENCE SUITE

ABSTRACT

This paper delineates the framework of Machine Intelligence Cyber Operations (MICO), which synthesises elements of cybersecurity with advancements in machine intelligence to augment Offensive Cyber Operations (OCO) substantially. The convergence of these two domains seeks to enhance cyber operations' effectiveness and strategic capacities, thereby contributing to a more resilient defence against contemporary cyber threats. This integration is pivotal in addressing the complexities of the modern cyber landscape and fortifying organisational preparedness in the face of evolving challenges. Our work focused on developing the Machine Intelligence (MI) suite to automate OCO's reconnaissance phase. A significant challenge faced was the lack of datasets to train the models. This was addressed by generating a 36278 x 27 dataset from general knowledge of network ports, services, and associated vulnerabilities. Our results show that the Multilayer Perceptron (MLP) regressor outperformed other models in 7 out of 7 metrics, with a Mean Squared Error of 0.0038 and R² score of 0.9999. Similarly, the MLP classifier achieved optimal performance among 5 other classifiers when evaluated with up to 8 performance metrics, with an accuracy of 1, F1-score of 1, and fastest prediction time of 0.035 seconds. The MLP Regressor and Classifier form the MI suite, predicting percentage vulnerability and suggesting possible attacks for maximum success. This MI suite will be integrated with other suites of the reconnaissance artefact, following the MICO framework, and analysed in our future study.

Keywords: Reconnaissance, MICO Framework, Machine Intelligence, Cyber Operations, MLP