I did a work that uses LRS to sign documents. LRS is a ring signatures model in which the feature of linking signatures made by the same signer is added. Using LRS two messages m1 and m2 are signed with key pairs (sk1, pk1) and (sk2, pk2), respectively, using the same N-group lambda (a group of public keys pki, i from 1 to N), and generate signature1 and signature2. A verifier that owns signature1, signature2 and lambda can check if sk1 is equal to sk2 without knowing any of sk1, sk2, pk1, or pk2.
I have a database composed of: N documents, N signatures, N public keys (1 per each participant that must have signed only 1 of the N documents).
One of the tasks I need to do now is to perform tests to verify that my public keys and their document signatures cannot be linked.
I read about NIST and understood that the tests performed are on already approved protocol algorithms and LRS is not one of them.
I thought about creating a large database and using some machine learning algorithm to try to detect a pattern but I don't know if this would be formally accepted if no pattern is detected.
Any idea how to formally proceed with this?
