R08: Slalom at the Carnival: Privacy-preserving Inference with Masks from Public Knowledge

Status: tracked
Year: 2024
Authors: Ida Bruhns, Sebastian Berndt, Jonas Sander, Thomas Eisenbarth
External source: Slalom at the Carnival: Privacy-preserving Inference with Masks from Public Knowledge

Source contribution

Carnival uses subset-sum-based masks to accelerate preprocessing for linear outsourcing.

PLLM perspective

Implement the exact published parameterized variant in an isolated comparison backend. Reproduce the reported Maverick attack before eligibility review.

Prior discussion identified this attack; no claim that our fresh full-ring masks instantiate Carnival.

This paper remains tracked. Named modules below are implementation targets, not finished PLLM components; no reproduction, integration, assurance, or matched benchmark is claimed.

The source's boundary remains binding: Maverick reports a privacy flaw for a Carnival instantiation. This is an attack-reproduction target, not an approved default or a blanket rejection of every variant.

PLLM modules and documentation

Target module identifiers: prepare.subset_sum, attack.carnival_projection. These identifiers describe scoped work unless this page explicitly links an implementation source. Dependencies: R07.

• Research backlog
• Reimplementation backlog
• Clean-room workflow

Backlog

This source is assigned to reimplementation priority 1, Delegated matrix-vector execution. Its source-registry order is 8. Follow current gates and unfinished work in the research backlog.