Multi-party Computation for Privacy-Preserving AML
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current AML detection processes in the financial sector are inefficient as they cannot combine personal information from different banks without violating data protection regulations, and existing multi-party computation solutions either require encryption or sharing of personal information, which are not compliant with regulations.
Innovation Solution
A computer-implemented method for parties A and B to calculate a Boolean function's result using secret inputs without sharing them, by generating random data items, dividing the function into subfunctions, and exchanging encrypted intermediate results, ensuring that only random data is shared across the network.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If banks share personal information about customers to enhance AML detection, then detection accuracy is improved, but data protection compliance deteriorates
Solution Approach 1:
The patent segments the AML detection process into multiple independent computation stages, where each party processes their own data locally and only exchanges non-sensitive intermediate results. This segmentation allows enhanced detection through combined analysis while maintaining data protection compliance by preventing direct sharing of personal information.
Solution Approach 2:
The patent introduces cryptographic protocols as intermediaries that enable computation on encrypted data without revealing the underlying personal information. These intermediaries facilitate accurate AML detection by allowing banks to query and analyze customer data from other banks while maintaining GDPR compliance through cryptographic guarantees.
2Reliability
If banks use encryption to protect customer data during exchange, then data protection compliance is maintained, but detection capability deteriorates
Solution Approach 1:
The patent applies preliminary cryptographic transformations to customer data before exchange, converting personal information into encrypted representations that retain computational utility. This preliminary action enables subsequent AML detection operations to be performed on encrypted data, maintaining both compliance and detection capability.
Solution Approach 2:
The patent changes the parameter representation of customer data from plaintext personal information to cryptographic encodings that preserve the essential characteristics needed for AML detection. This parameter transformation allows detection algorithms to operate on encrypted data with the same effectiveness as on plaintext, while maintaining data protection compliance.
3Reliability
If banks process AML detection independently without information sharing, then data protection compliance is maintained, but detection efficiency deteriorates
Solution Approach 1:
The patent merges the detection capabilities of multiple banks through cryptographic collaboration, allowing them to jointly analyze transaction patterns across institutions without consolidating personal data. This merging enhances detection efficiency by leveraging collective intelligence while maintaining compliance through distributed processing and cryptographic privacy guarantees.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
A computer-implemented method is carried out between a first party A and a second party B over a communications network, for calculating the result r of a Boolean function F having a first secret input a known to the first party A and a second secret input b known to the second party B, without either party sharing its secret input with the other party. The Boolean function is divided into separate subfunctions each of which outputs an intermediate result, and the intermediate results can be combined to provide the result r. Each party constructs from its secret two data items, one of which is not shared and the other of which appears to be random data and is shared. The subfunctions are constructed to enable each party to calculate all possible outputs for all possible values of the data item which was not shared by the other party. Using oblivious transfer and one-time-padding, each party can construct the final result r while only exchanging encrypted or random data with the other party and without ever sharing their secret data. The method has particular application in the field of anti-money laundering information sharing, where two or more parties are constrained from sending protected, personal or sensitive data to each other but wish to generate a result based on combined knowledge known individually to different parties.