A METRIC FOR THE APPLICATION OF HETEROGENEOUS DATASETS TO IMPROVE NEURAL NETWORKS IN CYBERSECURITY DEFENSE: A Quantitative Experimental Research Study

**My 2021 Quantitative Dissertation Study on the importance of EXTERNAL Cyber Threat Intelligence (CTI) in intrusion detection and prevention. The study's experimental results did not fully answer the hypothesis and lead to an overall rejection of it. Combining internal and external audit log security datasets did not provide measurable threat identification and error reduction using an Artificial Neural Network (ANN) architecture. The findings demonstrate that external or heterogeneous data alone was between 17 to 20 times more capable of reducing the neural network's error rate than using internal data and internal and external data in combination. Conversely, internal data demonstrates a less-than 1% improvement in its application to cybersecurity threat identification. The experiment's practicality provides quantified values for organizational economic planning and resource expenditures by an agency's leadership and Information Technology (IT) staff. The study's original contribution provides a data-enhanced neural network methodology that uses current or real-world threat vice outdated synthetic datasets. The experiment's results also identify a set of metrics to evaluate the quantified benefits of applying external or threat intelligence data to ongoing cybersecurity defense challenges.