Hi,
To find recording source/ device, initially I assumed that recorder information may be available in the metadata of recorded file. for this purpose I do audio recording on different mobile phones and upload on different social media platforms. After downloading the uploaded files, I analyse its metadata. From experimental results I reach to a conclusion that before uploading to social media platforms, we need to convert audio to video. The convertor replace the original metadata with its own new flags and values. So, it is not true to rely on metadata information and find recording source.
Now I am in search of some new techniques to find out some features/ parameters in the audio recordings that can discriminate different recording devices i.e. mobile phones (samsung, Iphone, infinix, Huwavie, redmi etc) and other recorders. Any help would be highly appreciated.
Haider.
Haider,
Metadata is only a part of recording device attribution. Anytime you upload an audio, video, or image, almost all websites "launder" (alter) the multimedia files. This can include altering metadata, re-quantization, changing bit-depth, resizing visual data, padding container files, etc.
That does not mean we cannot attribute the multimedia streams within the files to a recording device. It is frequently just a little more complicated, but still possible.
I recommend you check out the University of Colorado Denver's (UCD) National Center for Media Forensics (NCMF) research website for a large corpus of multimedia forensic research material, including recording device attribution. Full disclosure here, I am a part-time lecturer and thesis committee member at NCMF. I just published a technical note with a novel method you may wish to consider in your testing. The technical note is titled "Multimedia stream hashing: A forensic method for content verification." You can read the abstract and download the full paper at https://onlinelibrary.wiley.com/doi/abs/10.1111/1556-4029.15148. The paper was just an introduction of the method to the forensic community, not a complete validation study or anything. However, it covers the essential use of the technique. Just a side note: I am working on the image stream hashing validation study and hope to publish it late this year.
There are some other ways to attribute recordings to devices. Besides a good review of the NCMF research website, you may consider looking at Medex Forensics. https://medexforensics.com/. Again, full disclosure... I have no financial interest in their tool. The tools have some good foundational approaches to recording attribution to the devices. Much of their product is based on metadata, but there are other things they look at too. I believe Amped 5 may have some abilities too, but I am not sure.
Lastly, if you do not have the funds for automated tools or prefer to use your scripts, you know what they do to conduct your research/examinations (like me), then take a look at the NCMF research. I would use the research papers to expand your knowledge using the references in those papers with Google Scholar to find more data.
I hope this helps, and best of luck with your research.
Greg