Akademik Veri Yönetim Sistemi
15th International Conference on Information Security and Cryptography, ISCTURKEY 2022, Ankara, Türkiye, 19 - 20 Ekim 2022, ss.38-43
© 2022 IEEE.Significant advances in audio technology and enhanced functionality of editing tools have made it very easy to forge audio content. Many studies have been proposed in recent years to detect copy-paste, which is one of the most common types of forgery. In the copy-move fraud, the duplication of audio segments, the content is copied in a recording and moved to another place within the same recording. It is not always difficult to detect this. But things can get tricky due to the availability of sophisticated tools that can easily hide tampering without leaving any traces. Considering the importance of using audio signals in daily life, the necessity of an efficient audio verification system for accurate and reliable tamper detection is understood. The proposed method uses Mel spectrogram images obtained from the audio file and then converts the Mel spectrogram image to gray level images and divides it into overlapping sub-blocks and extracts features for each block via the Binary Gradient Model. Feature vectors are sorted lexicographically to group similar vectors together. The similarity between the blocks gives a clue about forgery. According to the given experimental results, we can say that the proposed scheme has better detection results even for tough attacks.