Quantitative study of zero‐amplitude sample padding in iOS AAC encoding: Consistency across devices and the impact of background noise

Zero-amplitude sample padding ("zero-padding") is automatically inserted during Advanced Audio Coding (AAC) encoding by iOS applications such as Apple Voice Memos, potentially affecting the reliability of forensic audio stream hashing by altering decoded audio data structure without impacting perceptual content. This study quantitatively analyzed 100 M4A (AAC) recordings from 11 iPhone devices across 5 distinct models under controlled high and low background noise conditions to examine the consistency of zero-padding across devices (RQ1) and the influence of background noise (RQ2). Pre-signal zero-padding was measured using validated methods across MATLAB, Pydub, and Py-audioread. Statistical analyses revealed significant variability in pre-signal padding (median: 2070.5 samples, Wilcoxon signed-rank test, p < 0.001, large effect size), exceeding the expected 1024-sample priming value described in Apple's documentation. Background noise significantly influenced padding, with higher noise associated with reduced pre-signal padding (Mann-Whitney U, p < 0.001, Cliff's delta = 0.244, 95% CI for median difference: -214.00, -61.00). Pre-signal padding measurements were consistent across tools, but post-signal padding showed tool-dependent variability. These findings provide a validated foundation for understanding zero-padding in iOS AAC encoding, informing the development of adaptive forensic verification methods to enhance the reliability of audio stream hashing in digital evidence analysis.