Is it possible to download corrupted torrent files, even with a download client builtin checksum verification?

Question

I downloaded some mkv/mp4 files with my new router (I can download directly from it, using transmission), and all of them seem to be corrupted: lags, strange artifacts when using vlc, crashes with mp4, etc.

When a downloaded part is corrupted, transmission warns you and downloads that part again. My question is: even after the transmission verification, is it possible that the downloaded file is still corrupt?

Solution 1

Yes, it is possible (although it might depend on the torrent client).

When I was having hard disk issues, my torrent client (Deluge) used to report files as downloaded 100%, when some parts of them were actually faulty. A forced re-check fixed that.

My educated guess is that the client downloaded the piece, checked it while it was stored in the RAM and then saved it to the hard drive (which makes sense having performance in mind). Something must have gone wrong while writing the data to the drive, thus corrupting the files.

If you force a re-check and the files still appear to be OK, chances are that it was a bad torrent or you have issues with your computer. A hash collision is theoretically possible, but highly unlikely in practice.

Solution 2

Saving data from RAM to disk is not error prone in any way. Only if there is a power problem or a serious disk issue which would be noticed very early due to errors in the OS.

Torrent poisoning is the only reliable answer I found so far.

Torrent poisoning is intentionally sharing corrupt data or data with misleading file names using the ?BitTorrent protocol. This practice of uploading fake torrents is sometimes carried out by anti-piracy organisations as an attempt to prevent the peer-to-peer (P2P) sharing of copyrighted content, and to gather the IP addresses of downloaders.

Source: http://en.wikipedia.org/wiki/Torrent_poisoning

Comments

[Synetech]

I’m sure collisions are not likely to be a problem since the hashes used are sufficiently large enough to make them rare. The question isn’t about the possibility of two files having the same hash, it is about whether the file could still be corrupt even after a corruption is detected. That is, whether the client will just ignore the error and leave it corrupt.

[Somebody still uses you MS-DOS]

This is an interesting theory. The re-check option exists on transmission. I'll need to redownload the torrent files though and test again. Thanks!

[Somebody still uses you MS-DOS]

Actually, the option is "Verify Local Data".

[Somebody still uses you MS-DOS]

You sir, hit the nail on the head. Re-checking did the work, it downloaded again a lot of parts. I don't know if your theory about the RAM is correct, but at least solved my problem. It's not the best thing to do (re-check every download since it takes a long time) but at least fixes the problems. Thanks!

[Olathe]

This isn't true. You can have a minor disk issue where only a few sectors go bad and then you can unluckily use those sectors for downloaded data.

[mateuszb]

I don't know how common the Torrent poisoning is, but I assume that HDD problems with sectors are less common. I might be mistaken though, I don't have any stats to back this up.

[Sergey]

I have similar problems with Deluge. Most of downloaded torrents are ok. But some of the torrents are shown as downloaded 100%, but after recheck they look downloaded less than 100%. But after download to 100% again, I'm getting into the same situation as it was at start. Is it possible somebody is seeding wrong data?

[Dennis]

@Sergey That's unlikely. Deluge should check the hashes of the downloaded blocks before accepting them.