SiliFuzz

Tool

First seen 7/18/2026

Last seen 7/18/2026

Evidence 12 chunks

NEIGHBORHOOD

No graph connections found for this entity yet. It may appear in future ingestion runs.

explore full graph →

RELATIONSHIPS

33 connections

Silent Data Corruption evaluates → 95% 2e

SiliFuzz targets silent data corruption issues in CPUs that do not cause immediate observable failures.

SiliFuzz Checker ← part of 100% 2e

SiliFuzz Checker is a component of SiliFuzz that runs multiple snapshots efficiently at scale.

F2XM1 Defect mentions → 95% 2e

SiliFuzz detected and reports the F2XM1 defect as one of the CPU defects found.

Cross-Pollination of Corpora implements → 90% 2e

SiliFuzz performs cross-pollination between corpora generated by different proxies.

libFuzzer uses → 100% 2e

SiliFuzz uses libFuzzer to fuzz XED and Unicorn as proxies for snapshot generation.

XED uses → 100% 2e

SiliFuzz fuzzes XED as a proxy to generate instruction sequences for testing CPUs.

Unicorn uses → 100% 2e

SiliFuzz uses Unicorn as a CPU emulator proxy for generating instruction sequences.

ifuzz uses → 100% 2e

SiliFuzz uses ifuzz to generate random instruction sequences for the corpus.

Fuzzing Corpus uses → 100% 2e

SiliFuzz accumulates a corpus of test inputs to execute on real CPUs at scale.

Snapshot uses → 100% 2e

SiliFuzz represents and executes test inputs as snapshots.

differential testing implements → 95% 2e

SiliFuzz cross-validates different CPU cores, which is a form of differential testing.

Coverage-Guided Mutation-Based Fuzzing depends on → 95% 2e

SiliFuzz heavily relies on coverage-guided mutation-based fuzzing advances.

Snapshot-Based Testing implements → 100% 2e

SiliFuzz uses snapshots as the core format for representing test inputs and replaying them on CPUs.

Proxy-Based Fuzzing implements → 100% 2e

SiliFuzz finds CPU defects by fuzzing software proxies such as simulators and disassemblers.

CPU Emulator as Proxy uses → 100% 2e

SiliFuzz uses CPU emulators as proxies to generate interesting instruction sequences.

CPU Electrical Defect evaluates → 100% 2e

SiliFuzz is designed to find and evaluate CPU electrical defects in production hardware.

SiliFuzz: Fuzzing CPUs by Proxy ← introduces 100% 1e

The paper introduces SiliFuzz as a work-in-progress system for finding CPU defects via proxy fuzzing.

Unicorn fuzz_emu_x86_64.c uses → 85% 1e

SiliFuzz applied libFuzzer to Unicorn in a setup similar to fuzz_emu_x86_64.c.

Nondeterminism Elimination implements → 95% 1e

SiliFuzz eliminates nondeterminism by tracing all snapshots and verifying deterministic replay.

End State Recording uses → 100% 1e

SiliFuzz records the end state of each snapshot execution to detect CPU defects.

Quarantine Pool uses → 90% 1e

SiliFuzz uses a quarantine pool to isolate and analyze defective machines.

Feedback Loop uses → 90% 1e

SiliFuzz includes a feedback loop between known bad machines and the fuzzing system.

Illegal Instruction Overshoot Defect mentions → 90% 1e

SiliFuzz detected a class of defects related to illegal instruction overshooting.

FCOS Miscomputation Defect mentions → 90% 1e

SiliFuzz detected a defect where FCOS produces incorrect results on some machines.

Missing x87 Data Pointer Update Defect mentions → 90% 1e

SiliFuzz detected a defect where the x87 data pointer is not consistently updated.

Corpus Distillation mentions → 85% 1e

SiliFuzz discusses corpus distillation as a future quality improvement strategy.

SiliFuzz references register scrambling as an expected effective snapshot mutation strategy.

gem5 mentions → 85% 1e

SiliFuzz mentions gem5 as a potential proxy for fuzzing microarchitectural models.

Simics mentions → 85% 1e

SiliFuzz mentions Simics as a potential proxy for fuzzing microarchitectural models.

RTL Design mentions → 90% 1e

SiliFuzz notes it does not have access to RTL design of target CPUs, distinguishing its approach.

Fuzzing Dictionary uses → 90% 1e

SiliFuzz uses ifuzz-generated instruction sequences as a dictionary for fuzzing Unicorn.

microarchitectural state uses → 85% 1e

SiliFuzz aims to improve testing by inheriting microarchitectural state from previously executed snapshots.

LLVMFuzzerTestOneInput XED Integration ← part of 90% 1e

The XED libFuzzer integration is part of SiliFuzz's snapshot generation pipeline.