MIT Trust Data Alliance building tomorrow’s smart city data systems
1 like280 views
The document discusses the MIT Trust Data Alliance which aims to build smart city data systems using federated data, distributed queries, and blockchain-based data control. It mentions analytics on encrypted data and using computation on encrypted data in a transparent and incorruptible manner. Examples provided include reducing exclusion and inclusion errors in Colombia's social programs by 21% each, detecting fraudulent accounts and hidden networks in bitcoin, and using little data to find targets through de-anonymization.
MIT Trust Data Alliance building tomorrow’s smart city data systems
- 1. MIT Trust Data Alliance building tomorrow’s smart city data systems “this will change everything” UN Secretary General’s Office Copyright 2019 Alex Pentland European Union Presidency Opening Keynote Speech • federated data with distributed query • blockchain-based data control • analytics on encrypted data Participants in trust.mit.edu alliance
- 2. transparent, incorruptable Output Inputs Blockchain Sandbox Computation Verify Identities, Computation Certifiedanswerstoverifiedentitites CertifiedOpenAlgorithms On-chain Off-chain CoreID and cheap! connection.mit.edu copyright Alex Pentland MIT 2019
- 3. How? Results: - 21% less exclusion error - 21% less inclusion error - at constant budget => 920k excluded poor now included Colombia ExclusionError Inclusion ErrorCopyright alex pentland 2018 all rights reserved
- 4. computation on encrypted data transparent, incorruptible, secure Output Inputs Blockchain Encrypted Data Verify Identities, Computation Certifiedanswerstoverifiedentitites CertifiedOpenAlgorithms CoreID Copyright alex pentland 2018 NO FIREWALL!!!
- 5. 4.53 4.54 4.55 4.56 4.57 4.58 4.59 4.6 4.61 x 10 5 4.53 4.54 4.55 4.56 4.57 4.58 4.59 4.6 4.61 x 10 5 nz = 11228155 Threshold =0.95 Real Time Community Analysis • Blocks are sub-communities • Some groups had NO internal direct money transfer! • We detect the entire “hidden network” around a single fraudulent account AI on encrypted data example: bitcoin fraud De-anonymization (“Identity merging”) Copyright alex pentland 2019, all rights reserved
- 6. ENDOR.com finding targets with very little data