paper 2025 ยท 19th International AAAI Conference on Web and Social Media (ICWSM)

MagnetDB: A Longitudinal Torrent Discovery Dataset with IMDb-Matched Movies and TV Shows

Scott Seidenberger, Noah Pursell, Anindya Maiti

Abstract

We introduce MagnetDB, a 15-year longitudinal dataset of torrent discovery data for movies and TV shows, matched with IMDb metadata. The release includes 4,936,257 magnet links and daily snapshots of over 130,000 unique torrents.

Key Results

  • Longitudinal tracking reveals content lifecycle and leak timing patterns in P2P networks
  • IMDb matching enables analysis of content traits associated with piracy demand
  • Public code and data support reproducible piracy and distribution research