System Immune: An AI-Driven Self-Healing Chaos Engineering Platform

Abstract

Modern distributed cloud systems are inherently complex, making them highly susceptible to failures such as service crashes, network latency, and database unavailability. Traditional recovery mechanisms depend heavily on manual intervention, resulting in increased downtime and operational costs. This paper proposes System Immune, an intelligent self-healing platform that combines Chaos Engineering with Reinforcement Learning to autonomously detect, diagnose, and recover from failures. Inspired by industry tools like Chaos. Monkey, the system introduces controlled fault injection while leveraging Q-Learning to determine optimal recovery strategies. The architecture utilizes Docker-based containerization for realistic failure simulation and a Flask-based backend for monitoring. Experimental evaluation demonstrates significant improvements in system resilience, reduced recovery time, and adaptive learning over repeated failure scenarios.The proposed system represents a step toward fully autonomous cloud infrastructure management.

Introduction

Cloud-native systems have transformed application deployment and scalability, but their distributed nature has increased challenges related to system reliability. Failures caused by hardware issues, software bugs, and network problems are common in modern systems. Traditional monitoring tools can detect failures but still depend on human intervention for diagnosis and recovery, leading to delays and downtime.

To improve resilience, Chaos Engineering was introduced as a proactive approach that intentionally injects failures into systems to test their robustness. Tools like Chaos Monkey simulate disruptions by randomly shutting down services. However, these tools mainly focus on testing failures and do not provide automated recovery.

The proposed system, called System Immune, combines Chaos Engineering with Artificial Intelligence to create a self-healing system inspired by the human immune system. The platform continuously monitors services, detects anomalies, and autonomously recovers from failures. It uses Reinforcement Learning, specifically Q-Learning, to improve recovery decisions over time and reduce dependence on human operators.

The related work discusses:

• Chaos Engineering for resilience testing through controlled failure injection.
• Reinforcement Learning in distributed systems for tasks such as resource allocation, load balancing, and fault recovery.
• Security and reliability standards such as the OWASP Top 10, which emphasize robustness and fault tolerance.

The study identifies a research gap because existing Chaos Engineering lacks intelligence, Reinforcement Learning lacks infrastructure integration, and few systems combine both approaches. System Immune addresses this by integrating AI-driven decision-making with real-world infrastructure testing.

The architecture of System Immune includes four main layers:

1. Infrastructure Layer – Uses Docker containers to isolate services and simulate realistic production environments.
2. Monitoring Layer – Continuously checks service health and detects failures like downtime, database disconnections, and latency issues.
3. AI Decision Layer – Uses Q-Learning to choose optimal recovery actions based on system states and rewards.
4. Execution Layer – Executes recovery actions automatically using Docker commands.

Additionally, the system provides a user dashboard for real-time monitoring and manual chaos injection. The platform also includes an attack generation and detection module that creates different security attack payloads and uses detection mechanisms to identify vulnerabilities accurately.

Conclusion

System Immune demonstrates how AI can enhance system resilience by enabling autonomous recovery. By combining Chaos Engineering with Reinforcement Learning, the system not only tests failures but also learns to fix them efficiently. This approach represents the future of intelligent cloud infrastructure.

References

[1] Chaos Engineering Literature [2] IEEE Research Papers [3] OWASP Top 10 [4] Reinforcement Learning Textbooks [5] Docker Documentation

Copyright

Copyright © 2026 Ms. Surabhi Chavhan, Nishad Palsapure, Mohd Ziyafat Abbas, Manas Bhajgawre, Mayur Chaple, Mohit Mene , Manish Khalsinghe. This 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.