<p>Meat adulteration compromises market fairness and poses risks to those with allergies or specific dietary needs. Ensuring meat authenticity is vital for food safety, public health, and consumer trust. This study presents a novel triple-hybrid deep learning architecture—CNN-GRU-LSTM—for accurate identification of pork adulteration in beef. The model is trained and evaluated using electronic nose (e-nose) sensor data collected from 420 samples across seven beef–pork mixture classes, obtained under controlled conditions using eight calibrated gas sensors and one temperature–humidity sensor (DHT22). The proposed architecture uniquely combines CNN to extract spatial patterns across multiple sensor channels, GRU to learn short-term temporal variations in volatile compound emissions, and LSTM to capture long-term temporal trends in volatile compound responses. By fusing the outputs of GRU and LSTM, the model generates a comprehensive temporal-spatial representation of sensor responses. The hybrid model achieved 99.91% mean accuracy over 5<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\times\)</EquationSource> </InlineEquation>5-fold cross-validation, significantly outperforming conventional classifiers. It also achieved outstanding results across all major evaluation metrics—precision, recall, F1-score, R², RMSE, and RPD—demonstrating the high efficiency and robustness of the proposed CNN-GRU-LSTM model, establishing it as a powerful solution for complex sequence modeling tasks such as meat adulteration detection.</p>