The evolving landscape of cybersecurity threats necessitates advanced detection systems to protect critical infrastructures and sensitive data. This research presents a blockchain-integrated security framework enhanced with machine learning for anomaly detection in manufacturing and energy systems. A hybrid ML model combining Long Short-Term Memory (LSTM) and Random Forest (RF) is employed to capture sequential attack patterns and classify threats with high precision. The system achieves an 18.9% increase in anomaly detection accuracy, 100% data integrity via tamper-proof logging, and reduces system downtime by 88.6%. Robustness is reflected through Byzantine fault-tolerant consensus (PBFT), adaptability to dynamic threats, and a 73.8% reduction in false positives. While the blockchain-based solution incurs a 14.8% increase in authentication time and a 5.8% drop in throughput, it significantly automates compliance and enhances auditability. The proposed model offers a promising approach for secure, intelligent, and resilient infrastructure protection in industrial environments.