Efficient Lattice-Based Ring Signature Scheme without Trapdoors for Machine Learning

Area of Science:

• Cryptography
• Lattice-based cryptography
• Machine Learning Security

Background:

• Machine learning (ML) and privacy protection are deeply intertwined, with ML systems being both targets and tools for privacy.
• Ring signatures (RS) offer cryptographic privacy, and lattice-based RS provides quantum resistance.
• Existing lattice-based RS schemes often rely on trapdoors, hindering computational efficiency due to hidden algebraic structures.

Purpose of the Study:

• To construct a novel, efficient, and secure lattice-based ring signature (RS) scheme for machine learning (ML) scenarios.
• To develop an RS scheme that avoids the computational overhead associated with trapdoor constructions.
• To enhance privacy protection in ML by leveraging advanced cryptographic techniques.

Main Methods:

• Utilized the Lyubashevsky collision-resistant hash function over lattices.
• Constructed a ring signature scheme based on ideal lattices.
• Employed the Fiat‒Shamir with aborts (FSwA) protocol for the construction.
• Analyzed security in terms of unconditional anonymity against chosen setting attacks (UA-CSA) and unforgeability with respect to insider corruption (EU-IC).

Main Results:

• Proposed a new lattice-based RS scheme without trapdoors.
• Achieved unconditional anonymity against chosen setting attacks (UA-CSA), a stronger security level than anonymity against full key exposure (anonymity-FKE).
• Satisfied unforgeability with respect to insider corruption (EU-IC).
• Demonstrated superior computational efficiency in signing and verification compared to existing schemes with similar security levels.

Conclusions:

• The developed lattice-based RS scheme offers enhanced privacy and security for ML applications.
• The absence of trapdoors and the use of ideal lattices lead to significant improvements in computational efficiency.
• This scheme is well-suited for ML environments requiring robust privacy protection and high performance.