Vibration-Based Detection of Hidden Drilling in High-Value Liquor

Hidden drilling provides a potential means of removing liquor from high-value bottles without visibly opening the package, posing a challenge for tamper detection. Here, we develop a vibration-based method for detecting drilling operations using vibration measurements acquired at the bottle-cap position and machine learning. An inertial measurement unit (IMU) mounted at the bottle-cap position was used to acquire vibration responses associated with drilling and typical transportation disturbances. Nine time- and frequency-domain features were extracted and used to train a radial basis function support vector machine (RBF-SVM) classifier. A total of 75 valid laboratory events were collected, including 35 Drilling events and 40 Transport events. Five-fold cross-validation achieved a mean accuracy of 98.33% (SD = 3.33%), while an independent test set yielded 100.00% Accuracy, Precision, Recall, and F1-score, with an area under the receiver operating characteristic curve (AUC) of 1.000. External validation under Small-Cart and bus transportation resulted in false-positive Drilling rates of 14.96% and 7.93%, respectively, reflecting the greater variability of vibration responses under real transportation conditions. In addition, RFID experiments evaluated the contactless identification of an RFID-equipped device under sealed-box conditions, with identification success rates varying across reading distances, reader orientations, and packaging conditions. These results support the feasibility of distinguishing hidden drilling from transportation-induced vibrations and provide a basis for package-level tamper detection without opening high-value liquor bottles.