RepuLink holds trust networks accountable with backward propagation
Backward endorsement penalties make trust networks accountable for misbehavior.
Trust and reputation systems are vital for reliable interactions in distributed networks, but current models rely solely on forward propagation of interaction-based trust signals. This leaves them unable to enforce accountability when negative interactions occur—endorsers of misbehaving nodes face no consequences. Additionally, new nodes with sparse interaction history suffer from the cold-start problem. In a new arXiv paper, Wenbo Wu and George Konstantinidis introduce RepuLink, a two-layer reputation model that solves both issues by coupling an endorsement network with an interaction feedback network.
RepuLink integrates two concurrent backward propagation mechanisms: Backward Endorsement Penalty Propagation (BEPP) recursively penalizes endorsers of misbehaving nodes, while Backward Endorsement Reward Propagation (BERP) rewards endorsers of well-performing nodes. This creates a positive interaction feedback loop and makes endorsements accountable. The endorsement layer also provides explainable, endorser-weighted trust initialization for new nodes, addressing cold-start. Experiments on real-world datasets show RepuLink outperforms representative trust propagation baselines across all four evaluation metrics in both interaction-only and full two-layer settings, while preserving comparable efficiency.
- BEPP recursively penalizes endorsers of misbehaving nodes, enforcing accountability for propagated trust signals.
- BERP rewards endorsers of well-performing nodes, incentivizing positive behavior and creating a feedback loop.
- RepuLink outperforms existing trust propagation baselines across four evaluation metrics on real-world datasets.
Why It Matters
Makes decentralized trust systems accountable, solving cold-start for new nodes with explainable initialization.