Search papers, labs, and topics across Lattice.
To eliminate Personally Identifiable Information (PII) exposure in financial surveillance, this work executes fraud triage directly over topologically anonymized latent transaction spaces. The approach couples iterative rounds of unsupervised candidate pruning with targeted supervised classification to isolate malicious behavior purely through geometric manifold properties. The resulting system achieves ultra-low latency detection while strictly preserving user privacy across institutional boundaries.
Financial fraud can be caught directly from the manifold: topological latent spaces retain enough geometric signal to snipe malicious transactions at ultra-low latency without ever exposing underlying PII.
Working entirely on topologically anonymized embeddings, we perform fraud detection using iterative rounds of unsupervised filtering followed by supervised sniping. The result is an ultra-low latency privacy--preserving triage that allows institutions to flag suspicious activity without compromising Personally Identifiable Information.
