Assessing APT Detection Using Financial AI and Machine Learning: Can Greater Accuracy Be Achieved? (CASE STUDY)

EasyChair Preprint 14872

Abstract

Having been one of the most complex and serious challenges in the cybersecurity space, APTs are cleverly designed, therefore managing to evade various traditional mechanisms of detection. Since machine learning has emerged as one of the most effective tools in cybersecurity, this article considers a review on the effectiveness of ML-based techniques in detecting APTs and explores whether superior accuracy is achievable. We review various ML models to discuss strengths and weaknesses in APT detection along with the enhancements being done in the area of data quality and feature selection for the betterment of the detection. We comparatively review the existing approaches to give an insight into the potentials of ML in improving the accuracy of APT detection. The results reflect that though ML offers promising enhancements, model selection, training data, and, above all, the constantly changing nature of APTs require careful consideration in order to achieve superior accuracy with consistency. This has very much significant implications for cybersecurity practices as organizations are eager to implement more robust and reliable methods against these stealthy threats.

Keyphrases: APT detection, Advanced Persistent Threats (APTs), Cybersecurity, Model Interpretability, Reinforcement Learning, anomaly detection, feature engineering, machine learning, supervised learning, unsupervised learning

@booklet{EasyChair:14872,
  author    = {Adeoye Ibrahim},
  title     = {Assessing APT Detection Using Financial AI and Machine Learning: Can Greater Accuracy Be Achieved? (CASE STUDY)},
  howpublished = {EasyChair Preprint 14872},
  year      = {EasyChair, 2024}}