Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{204055,
        author = {Mohit and Ms. Sarika Madavi},
        title = {A Systematic Review of Hybrid Machine Learning and Agentic AI Frameworks},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {1},
        pages = {596-604},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=204055},
        abstract = {The rapid evolution of cyber threats has created serious challenges for conventional cybersecurity systems. Traditional rule-based intrusion detection mechanisms and standalone machine learning models often struggle to identify sophisticated attacks such as zero-day exploits, polymorphic malware, ransomware, distributed denial-of-service attacks, and advanced persistent threats. Recent advancements in Artificial Intelligence have introduced Agentic AI as a new paradigm capable of autonomous reasoning, adaptive learning, contextual understanding, and real-time decision-making. The integration of Machine Learning (ML) and Agentic AI has emerged as a promising approach for developing intelligent and self-adaptive cyber defense systems.
This review paper presents a comprehensive analysis of hybrid ML and Agentic AI frameworks for autonomous cyberattack detection and response. The study critically examines two major research frameworks focused on intelligent cyber defense through the fusion of predictive analytics and autonomous agent-based orchestration. The paper explores various deep learning architectures, including Convolutional Neural Networks, Recurrent Neural Networks, Long Short-Term Memory networks, ensemble learning models, reinforcement learning mechanisms, and goal-oriented autonomous agents used in modern cybersecurity ecosystems.
The review discusses system architectures, anomaly detection models, contextual risk evaluation techniques, adaptive policy optimization mechanisms, and reinforcement learning-based mitigation strategies. Furthermore, the paper analyzes performance metrics including detection accuracy, precision, recall, F1-score, false positive reduction, mitigation latency, and scalability in enterprise, cloud, and IoT environments. The comparative analysis demonstrates that hybrid ML-Agentic AI systems significantly improve threat intelligence, adaptive response capabilities, and autonomous mitigation efficiency compared to traditional ML-only and rule-based approaches.
The paper also highlights recent advancements such as explainable AI, federated learning, retrieval-augmented generation, blockchain-enabled trust systems, TinyML security architectures, and self-healing cybersecurity infrastructures. Finally, major research challenges, ethical concerns, scalability limitations, adversarial vulnerabilities, and future research opportunities are critically discussed to support the development of next-generation intelligent cyber defense ecosystems.
The speedy evolution of cyber threats has created severe demanding situations for conventional cybersecurity structures. traditional rule-based intrusion detection mechanisms and standalone system gaining knowledge of models often war to pick out sophisticated attacks along with 0-day exploits, polymorphic malware, ransomware, disbursed denial-of-carrier attacks, and advanced persistent threats. current advancements in synthetic Intelligence have delivered Agentic AI as a brand-new paradigm capable of autonomous reasoning, adaptive gaining knowledge of, contextual information, and actual-time selection-making. the integration of gadget mastering (ML) and Agentic AI has emerged as a promising method for growing intelligent and self-adaptive cyber protection structures.
This review paper affords a comprehensive evaluation of hybrid ML and Agentic AI frameworks for independent cyberattack detection and reaction. The examine severely examines foremost research frameworks centered on sensible cyber defense via the fusion of predictive analytics and self-sufficient agent-based orchestration. The paper explores diverse deep studying architectures, which includes Convolutional Neural Networks, Recurrent Neural Networks, lengthy short-term reminiscence networks, ensemble gaining knowledge of fashions, reinforcement studying mechanisms, and aim-oriented independent dealers used in modern cybersecurity ecosystems.
The review discusses device architectures, anomaly detection fashions, contextual danger evaluation strategies, adaptive policy optimization mechanisms, and reinforcement studying-primarily based mitigation strategies. moreover, the paper analyzes overall performance metrics such as detection accuracy, precision, recollect, F1-score, fake advantageous discount, mitigation latency, and scalability in company, cloud, and IoT environments. The comparative evaluation demonstrates that hybrid ML-Agentic AI structures considerably improve danger intelligence, adaptive response abilities, and self-sustaining mitigation efficiency as compared to traditional ML-only and rule-based tactics.
The paper additionally highlights recent advancements such as explainable AI, federated getting to know, retrieval-augmented generation, blockchain-enabled trust structures, TinyML safety architectures, and self-restoration cybersecurity infrastructures. ultimately, important research demanding situations, ethical issues, scalability boundaries, antagonistic vulnerabilities, and destiny studies possibilities are severely discussed to aid the improvement of subsequent-generation shrewd cyber defense ecosystems.},
        keywords = {Machine Learning, Agentic AI, Autonomous Cyber Defense, Intrusion Detection Systems, Deep Learning, Threat Intelligence, Reinforcement Learning, Adaptive Security, Cyberattack Detection, Explainable AI.},
        month = {June},
        }