Depending on the exact use case, you could decode the mp3 files, align both that they are in-phase (don't know if that's the correct word, mainly finding the point where both signals have the same waveform, works good for digital stuff, and less good if you have different recordings from tape), then phase invert one and adding them up (or simply subtract one of the other..). then calculate the peaks, if these are below a certain threshold (around < -50dB for audio from CD-Masters is a good value, but depends on the audio and encoding) chances are good it's the same audio. getting the audio programmatically in-phase could be tricky it think, again, depending on the audio / conversions / encodings. maybe that helps.