Detecting APT-Induced Network Anomalies with AI: A Hybrid Statistical–Deep–Graph Framework
Preprint
- 26 August 2025
- preprint
- Published by MDPI AG in Preprints.org
Abstract
Advanced Persistent Threats (APTs) are among the most dangerous and sophisticated types of cyberattacks, capable of infiltrating enterprise networks and staying hidden for long periods while stealing data or causing damage. Unlike random cyber threats, APTs use stealth, persistence, and adaptive tactics to evade traditional defenses. This article examines the unusual activities caused by APTs within network environments, focusing on how Artificial Intelligence (AI) can be used to detect these anomalies effectively. We start by explaining the basics of APTs, their lifecycle, and features that make detection dif-ficult. Then, we review a wide range of anomaly detection methods, especially those driv-en by AI. A detailed mathematical framework for identifying anomalies is presented, in-cluding hypothesis testing, change-point detection, probabilistic models, and graph-based learning techniques. The development section describes AI-based detection methods, with mathematical details and pseudo-code. Results from controlled experiments and bench-marks like UNSW-NB15, CIC-IDS2017, and CTU-13 are analyzed, followed by a discus-sion of strengths, limitations, and future research directions. The article wraps up with strategies for deployment and a plan for real-world implementation.Keywords
All Related Versions
This publication has 19 references indexed in Scilit:
- Machine Learning and Deep Learning Methods for CybersecurityIEEE Access, 2018
- A survey of network anomaly detection techniquesJournal of Network and Computer Applications, 2016
- An empirical comparison of botnet detection methodsComputers & Security, 2014
- A survey on concept drift adaptationACM Computing Surveys, 2014
- An Overview of IP Flow-Based Intrusion DetectionIEEE Communications Surveys & Tutorials, 2010
- A detailed analysis of the KDD CUP 99 data set2009
- Internet traffic classification using bayesian analysis techniquesACM SIGMETRICS Performance Evaluation Review, 2005
- Estimating the Support of a High-Dimensional DistributionNeural Computation, 2001
- Bro: a system for detecting network intruders in real-timeComputer Networks, 1999
- Term-weighting approaches in automatic text retrievalInformation Processing & Management, 1988