SEISMO: System for tidlig deteksjon av mistenkelige markedsplass operasjoner (SkatteFUNN)

Offentlig prosjektsammendrag

Online fraud is a large societal challenge, and detecting fraud before it takes place is particularly hard. C2C platforms and marketplaces, such as Tise, are uniquely vulnerable due to limited detection resources, high user transaction/interaction volumes, and rapid transaction processes that appeal to scammers. The innovation effort will focus exclusively on C2C interactions, propose fraudulent interaction methods that by design are privacy-sensitive and avoid user profiling, have access to a large peer-to-peer Norwegian interaction pattern dataset, and supplement platform records with data based on experiences of fraudulent interactions. Six research challenges are highlighted to advance the research front: * Sequence pattern mining and interaction classification * Causal Bayesian networks and classification * Intent classification with lightweight transformer model * User interaction event collection * Streaming-based aggregation of event data * Data monitoring system