Search papers, labs, and topics across Lattice.
This paper introduces a novel method for verifying the authenticity of audio in short viral videos by focusing on external consistency with trusted news recordings rather than detecting internal inconsistencies. By formulating a likelihood ratio test based on the alignment of audio queries with reference recordings, the authors demonstrate that their approach significantly outperforms traditional methods that rely on MFCC features and Euclidean distance. The key result shows that this dual alignment likelihood ratio test is not only faster but also offers greater robustness and explainability in verifying audio integrity.
Verifying audio authenticity through a dual alignment likelihood ratio test reveals a faster and more robust alternative to traditional tampering detection methods.
This paper explores a way to verify that audio has not been maliciously tampered in a specific context: short viral videos taken from news recordings. Rather than trying to detect artifacts of tampering (internal inconsistency), we focus on positively verifying a query against a trusted source such as a news recording (external consistency). We propose a method for cross verifying a short audio query against a reference recording from which it was taken. Our approach is to define two hypotheses (non-tampered vs tampered), calculate the most likely alignment between query and reference for each hypothesis, and then perform a likelihood ratio test on the two alignments. We show that this method is fast to compute, much more robust than using MFCC features with Euclidean distance, and has the key benefit of explainability. Our cross verification approach provides an alternative perspective and complementary tool to existing tampering detection methods.
MIT CSAIL