Systems and methods for verifying decentralized federated data using influence evaluation

摘要

The disclosed computer-implemented method for verifying decentralized federated data using influence evaluation may include (i) calculate an influence score for each of a group of data instances, (ii) rank the data instances based on the influence scores, (iii) determine an anomaly score for each of the ranked data instances, (iv) select the ranked data instances with the highest anomaly scores as containing potentially malicious data, and (v) perform a security action that protects against the potentially malicious data. Various other methods, systems, and computer-readable media are also disclosed.