This study explores the potential of hyperspectral imaging (HSI) combined with machine learning for the non-destructive detection of ghoore (unripe grape) adulteration in brown sumac, a medicinally and economically valuable spice vulnerable to quality degradation. Samples with adulteration levels of 5%, 20%, 35%, 50%, and 100% were analyzed. Hyperspectral images were acquired and processed using spatial segmentation and Savitzky–Golay filtering to extract informative spectral features. Classification models including Random Forest (RF), Support Vector Machine (SVM), Extreme Gradient Boosting (XGBoost), and Artificial Neural Network (ANN) were employed for both binary (pure vs. adulterated) and six-class (specific adulteration levels) classification. The SVM model achieved the highest accuracy, with 99.00% for binary classification and 94.55% for six-class classification. Key discriminative features identified through RF and XGBoost analysis included phase-related components, fractal dimension, and the area under the curve (AUC) in the 700–900 nm spectral range. The results demonstrate that the integration of HSI and machine learning enables a rapid, non-destructive, and reliable method for detecting sumac adulteration, offering significant potential for food quality assurance applications.