The problem of Internet security is becoming more and more prominent, and the traditional single encryption algorithm has the situation of insufficient security in the face of complex threat scenarios. Therefore, a reliability detection method for Internet of Things cloud security based on Secondary Encryption Algorithm (SEA) is studied and combined with the abnormal data detection mechanism of Gaussian mixture model to improve the protection capability of the system under diversified attack scenarios. The experimental results show that SEA is significantly better than the traditional algorithm in multi-dimensional performance index. The defense success rate of SEA is more than 99%, and the accuracy score of anomaly detection is 9.5. The encryption speed is 890 ms/MB, and the data processing throughput reaches 2500 bars/second. When the nodes are expanded to 100, the efficiency decreases by only 5%, which has strong stability and adaptability. In summary, the SEA system can effectively improve the efficiency and accuracy of IoT cloud security reliability detection, and provide a reliable solution for security protection in a highly dynamic environment.