Optimal ordering of oracle queries by expected information gain

Determine whether ordering a fixed set of oracle queries for locked netlists by expected information gain improves on random queries or satisfiability-based query selection.

Background

The paper compares uniformly random oracle queries with queries selected by a satisfiability-based attack. It shows that chosen queries can reduce the surviving key version space more rapidly per query, whereas random queries are cheaper and can outperform chosen queries when sufficiently many are available.

The paper does not investigate how to order a predetermined set of queries using the greedy expected-information-gain criterion developed earlier for modified-Haar coefficients. Consequently, it leaves unresolved whether such ordering provides an advantage over either random selection or satisfiability-based query generation in the locked-netlist setting.

References

And while Section~\ref{sec:adversarial} compares random queries against the patterns a satisfiability-based attack selects, it does not order a fixed set of queries, so whether ordering by expected information gain improves on either choice is still open.

— A Version Space Approach for Digital Circuit Analysis  (2609.00609 - Thornton, 1 Sep 2026) in Section 5.3, “Scope and What Remains” (label: sec:scope)