VolodyA! V Anarhist wrote: > Brian Mearns wrote: >> On Thu, Sep 10, 2009 at 6:54 AM, SmallSister development >> <smallsis...@xs4all.nl> wrote: >>> Brian Mearns wrote: >>>> I understand the basic architecture of Freenet and how it protects >>>> contributors, but I'm concerned that the files themselves might >>>> contain identifying information about the source of the file. In >>>> particular, for audio files such as MP3, Ogg Vorbis, and Flac. Does >>>> anyone have any information on making sure there are no >>>> "fingerprints" left on these files when posting? >>> There is no way to be certain. There could be a future algorithm piecing >>> together just enough bits from a file to identify you. Word frequency >>> analysis has been used to discern likely writers of a plain text file. >>> Yes, you can avoid stupid mistakes by checking for "known signatures" >>> and removing them, but that is something that would be different for >>> different file types (ogg would be somewhat different from mp3; but doc >>> and pdf require totally different approaches.) It would be useful to >>> have tools that can identify and remove watermarks and Freenet would be >>> a great place to publish them (with source code please!) >>> >> Bugger, I didn't even think of steganographic watermarks (shame on >> me), I was thinking more in terms of meta data known to be stored with >> various file formats (like ID3 tags), but this seems to be more >> complicated than I expected. Thanks for the insight. > > For audio and video the best approach could be to take the streams and remux > them into the same container type, then to copy only the metadata which the > user > knows about (for example some audio files have autogain settings, etc). > > You would still have a potential for watermarking, as Peter already pointed > out, > and these watermarks will become more and more common as Internet Music Shop > software develops (this is not to assume that internet shops is the only > source > of watermarked audio, other groups can watermark a file, this is most > dangerous > in countries where freenet is illegal(*)). > I have heard noticed that multiple rips of the same CD can produce different .wav files; some of the differences are caused by the CD-player firmware and some caused by scratches/defects on the CD. Add some compression artefacts of the mp3 encoder (subtle algorithm differences between implementations) and you're very likely able to identify each mp3 uniquely. Even re-encoding may be detectable. NOTE: this is all without explicitly inserting a watermark.
Remuxing and detection and removal of "steganographic" watermarks is likely to bring you 95% of the way; which could be good enough. _______________________________________________ Support mailing list Support@freenetproject.org http://news.gmane.org/gmane.network.freenet.support Unsubscribe at http://emu.freenetproject.org/cgi-bin/mailman/listinfo/support Or mailto:support-requ...@freenetproject.org?subject=unsubscribe
