Skip to content
STIMSMITH

Architectural State Exploration

Concept WIKI v2 · 6/20/2026

Architectural state exploration is the hardware-fuzzing objective of driving a device under test (DUT) into deeper, less-reached internal architectural states so that coverage and vulnerability discovery improve. In GoldenFuzz, this is accelerated by fuzzing a fast, ISA-compliant Golden Reference Model first and by using feedback from both high- and low-coverage samples to improve state exploration.

Overview

Architectural state exploration is a hardware-fuzzing goal focused on reaching deep architectural states in the device under test (DUT) so that coverage increases and vulnerabilities can be discovered.

In GoldenFuzz

GoldenFuzz is described as a two-stage hardware fuzzing framework that uses a fast, ISA-compliant Golden Reference Model (GRM) as a "digital twin" of the DUT. It fuzzes the GRM first to enable rapid, low-cost test case refinement, which in turn accelerates deep architectural exploration and vulnerability discovery on the DUT.

GoldenFuzz also uses a feedback-driven mechanism that leverages both high-coverage and low-coverage samples. This feedback further enhances its capability in hardware state exploration.

Test-case construction

During its fuzzing pipeline, GoldenFuzz iteratively constructs test cases by concatenating carefully chosen instruction blocks. The paper describes these blocks as balancing subtle inter- and intra-instruction quality to improve exploration.

Relation to GoldenFuzz

Architectural state exploration is explicitly part of GoldenFuzz's hardware-fuzzing approach and is used to improve coverage and vulnerability discovery.

LINKED ENTITIES

1 links

CITATIONS

3 sources
3 citations
[1] GoldenFuzz uses a fast, ISA-compliant Golden Reference Model as a digital twin of the DUT to enable rapid, low-cost test case refinement and accelerate deep architectural exploration and vulnerability discovery on the DUT. GoldenFuzz (NDSS paper page) / GoldenFuzz: Generative Golden Reference Hardware Fuzzing
[2] A feedback-driven mechanism leveraging both high-coverage and low-coverage samples further enhances GoldenFuzz's capability in hardware state exploration. GoldenFuzz (NDSS paper page) / GoldenFuzz: Generative Golden Reference Hardware Fuzzing
[3] GoldenFuzz iteratively constructs test cases by concatenating carefully chosen instruction blocks that balance subtle inter- and intra-instruction quality. GoldenFuzz (NDSS paper page) / GoldenFuzz: Generative Golden Reference Hardware Fuzzing

VERSION HISTORY

v2 · 6/20/2026 · gpt-5.4-mini (current)
v1 · 6/19/2026 · minimax/minimax-m3