NETWORK ATTACKS AND THEIR CLASSIFICATIONS: A SYSTEMATIC REVIEW OF THREATS, DETECTION MECHANISMS, AND DEFENSE STRATEGIES
Annotatsiya
Growing network complexity and IoT proliferation have expanded the attack surface for malicious actors, demanding more robust intrusion detection solutions. This paper presents a systematic review of seven major network attack categories — DoS/DDoS, Man-in-the-Middle, IP Spoofing, ICMP-based attacks, Lateral Movement, SQL Injection, and Telnet exploitation — mapped to OSI layers and MITRE ATT&CK techniques. Seven ML and DL detection models (SVM, Random Forest, LSTM, GRU, CNN, GNN) are evaluated across KDD Cup 1999, NSLKDD, UNSW-NB15, and BoT-IoT datasets. LSTM achieves the highest accuracy of 99.30% with a 0.68% False Alarm Rate; tuned SVM yields 99.12% accuracy. Four real-world case studies — Mirai, SolarWinds, Stuxnet, and Log4Shell — validate the taxonomy. Results show that a hybrid approach combining anomaly-based ML, behavioral analytics, and graph-based detection offers the most effective defense. Future directions include adversarial robustness, federated learning, and zero-trust enforcement.
Hali tarjima qilinmagan