Combining auto encoders and hybrid cellular automata provides a novel way to identify anomalies in structured data in the field of anomaly detection. Dimensionality reduction and extracting the features is one of the know expertise of Auto encoders. Auto encoders compress and rebuild data into a lower-dimensional space, emphasizing anomalies with more reconstruction errors for non-typical data. Simultaneously, hybrid cellular automata provide a layer of spatial dynamics modelling by controlling the evolution of data through rule-based interactions among neighboring units. The hybrid methodology has the advantages of both systems: the local interaction modelling of hybrid cellular automata and the pattern recognition powers of auto encoders. This novel method is obtaining an accuracy of 99.83% which is promising, when comparing with the baseline methods. The integration of hybrid cellular automata instead of linear cellular automata has improved the overall accuracy and performance of the proposed classifier. We have used various parameters like F1 Score, precision, recall and ROC for measuring the effectiveness of the proposed system. The accuracy reported for anomaly detection in financial tractions is 99.63 and video surveillance is 99.96.