STDINFO PROTOCOL DESCRIPTION


Secure multiparty computation to infer transmission risk from 2 sexually transmitted diseases (STDs) test reports.

The mathematical problem statement is: compute a function F(a,b) that takes input a from Alice and b from Bob, without revealing b to Alice nor a to Bob

Bob encrypts b and send enc(b) to Alice. Alice computes F(a,enc(b)) sends it to Bob. Bob can compute F(a,b) without knowing b, because F is a special function that allows that using garbled circuits, finite field arithmetic, or Shamir secret sharing.

Sample problem set up

Alice and Bob have 2 private strings. A = 010001 and B = 010000

Compute bitwise XOR and modulo-2 sum. 0+0+0+0+0+1 = 1. If the result of this computation is 1, both partners cannot have safe sex. If the result is 0, both partners can have safe sex without transmission risk.

Bit strings represent medical reports. Bitwise XOR and modulo-2 sum returns 1 if A has a sexually transmitted disease that B doesn't, and True (0) if both can have safe sex because they either have no diseases or share the same diseases.

STDinfo protocol underlying working principle

There are 3 parties: A (Alice), B (Bob), and C (Charlie). A and B have private data 'a' and 'b'.

A generates shares s1,s2,s3 from 'a' with the property that any 2 can be used in a public function g() to get a. This property is called Shamir secret sharing. B similarly generates shares t1,t2,t3 from 'b'.

A sends s2,s3 to B and C respectively. B sends t1,t3 to A and C respectively.

A calculates XOR(s1,t1) and sends it to B. B calculates XOR(s2,t2) and sends it to A.

A and B calculate XOR(a,b) using XOR(s1,t1) and XOR(s2,t2) because XOR(s1,t1) and (s2 XOR t2) are shares of XOR(a,b). (from g(s1,s2)=s). This is called 2-reconstruction property.

Thus, XOR (a,b) is calculated without sharing a with Bob nor b with Alice. Similarly, modulo-2 is calculated.

Lasso is a regression algorithm that ignores features having low correlation with a target. This was used to find important features from a combination of 2 private datasets without the parties revealing datasets to each other.

STDinfo protocol has applications in semiconductor fab optimization, healthcare data analytics, and privacy for large language models.

MPyC: Library in Python for secure multiparty computation

STDinfo: Github repository for ONITO's Python implementation of secure multiparty computation to find disease transmission risk from medical test reports








ONITO DEX