Idea for file/checksum test

[Samuel T Cogley]

The thread that @The Computer Audiophile just closed got me thinking.

 

What is the shortest duration a WAV file would have to be to demonstrate the "checksums identical, but sounds different"?

 

If the files were small enough, we could actually look at the individual bytes of the files (no checksums) to confirm they are or aren't identical.

 

But this all really hinges on how small the file can be and still show the purported issue.

 

Just throwing it out there.

[mansr]

I made the files by simply adding noise and then using a chosen prefix attack on md5. The collision is produced by adding junk, which is ignored by playback software, to the end of the files. This is the software I used: https://github.com/cr-marcstevens/hashclash It took about 5 hours on an old 6-core Xeon.

[yamamoto2002]

Following two wav files have the same MD5 value (but SHA-1 or other checksum numbers are different).

 

File size is 704 bytes each, Sound duration is 0.003 seconds.

 

SameMD5Wav.zip