Know what you are about to download
Every file sharing network has them: the film that turns out to be a different film, the album that is a program, the release that wants a password nobody has. Classic eMule leaves you to find out after the download. eMule Qt weighs the evidence beforehand and tells you what it thinks — and why.
A Confidence column
Search results, indexer results and the download list each have a Confidence column. Every file gets a fake score from 0 to 100 and one of six verdicts:
| Verdict | What it means |
|---|---|
| Genuine | Nothing against it, and something for it: users rated it good, or many Kad publishers agree on what the file is. |
| Looks good | Nothing, or very little, against it. |
| Caution | A score of 25 or more. Shown with the number. |
| Suspect | A score of 50 or more. |
| Likely fake | A score of 75 or more. |
| Spam | The spam filter already counts it as spam. |
The column is coloured and sorts worst-first or best-first, so the doubtful results of a search can be pushed out of sight with one click. The web interface shows the same column, and the REST API and the MCP tools return the verdict with its reasons.
What it looks at
- Names that disagree — one file shared under names that describe different content, or as different kinds of file (a video here, an archive there). The words you searched for are ignored, since they are in every name.
- Disguises — a program or an archive named like a media file, or a published file type that does not fit the extension.
- The first bytes — once a download has its beginning, the content is compared with what the extension claims.
- Media data that cannot be true — a length or bitrate that does not fit the size, or an artist, album and title that appear in none of the names.
- What other users say — ratings, comments, Kad notes that call a file a fake, and the spam filter’s own rating.
- Hashes that disagree — sources reporting different AICH hashes for the same file.
- Locked releases — a Usenet release listed as password protected with no password supplied.
Each reason counts once, and popularity alone proves nothing: a popular fake has many sources too.
It tells you why
Hover a verdict and the tooltip gives the score and one line per reason — “It is a program named like a media file”, “Sources disagree on the AICH hash”. You are shown the evidence, not just a colour.
It remembers
The names a file has gone by are kept across restarts, so a fake that shows a different name in every search is still recognised as one file with many stories. The memory is bounded: entries nobody has seen again for half a year are dropped.
A warning, never a filter
The detector hides nothing and blocks nothing. A file marked “Likely fake” stays in the list and can be downloaded like any other. The decision is yours.
[tokens], regular expressions under [regex]. Edit it with any text editor; it is read when the program starts.