Metadata Independent Hashing for Media Identification & P2P Transfer Optimisation

Efficient swarming behaviours within peer-to-peer networks are hindered by imprecise or incorrect metadata content. Once published, metadata corrections can only be effected by a complete republish/swarm recreation or for each peer to manually make corrections (causing them to leave the swarm, decreasing performance). This work presents an approach which enables a swarm to collaboratively upgrade embedded data to reflect changes in metadata, and to identify additional candidates which contain differing metadata but a correct payload. Swarm degradation due to peer drop-off resulting from edits is eliminated, and additional peers can be identified in a fully automated fashion, increasing swarm lifetime and performance. Arising from this metadata abstraction, automatic purification can be realised in situations where multiple incomplete/incorrect versions are available within one or more unconnected swarms. Variations associated with a content set are processed associatively using a knowledge discovery rule set to extrapolate a canonical tag set, which can also be reinforced using data from external corpora. After any update, these changes can again be automatically disseminated in a peer-to-peer swarm. The system presented enables context-aware P2P data transfers which abstract metadata optimally, while also maximising swarm size and enabling cataloguing of content. A proof-of-concept implementation is presented, and its impact on swarm purification/ optimisation is evaluated.