The Dark Side of Threshold Cryptography

Abstract

It is typical for a cryptographic technology to be useful in its primary goal and applications, yet to exhibit also a dark side, namely to allow abuses in some other situations. Examples are subliminal channels in strong (randomized) signature schemes, employing authentication for encryption, kleptography exploiting strong randomness, etc. Threshold cryptography was introduced to realize better security and availability. However, its “dark side” has never been addressed seriously. We investigate some possible abuses of threshold cryptography which result from users not possessing the entire private key due to threshold splitting. This is a deficiency which can hurt de-commitment in procedures like “contract signing” and nullify non-transferability properties. To attempt solving the problem, one may suggest to assure that the user has full control of his private key via a zero-knowledge confirmation. However, advances in cryptography itself defeat this, since the Completeness Theorem for secure computations implies that servers in possession of shares of a key can answer on behalf of the “virtual holder” of the entire private key, without compromising the secrecy of their shares. We are then forced to look at more physical limitations of the setting. We propose a notion we call Verifiable Secret Non-Sharing (VSNS) where we can replace the strong (i.e., less realistic) physical isolation assumption (namely, a Faraday cage) with a more realistic timing assumption. We then introduce a new class of “combined software engineering and cryptography” adversarial capability, which employs software preprocessing and cryptography in breaking all previous suggestions to our problem. It seems that the adversary is so powerful that we have to rely on certain tamper-resistant device to block it. We show how to prevent a malicious participant from compromising the secrecy of the provers’ secret keys in this case. Our treatment is a step towards a model of computing with trusted and non-trusted tamper-proof (black box) elements.

The whole is more than the sum of its parts