HeteroFuzz

Tool

First seen 7/26/2026

Last seen 7/26/2026

Evidence 16 chunks

NEIGHBORHOOD

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

17 connections

HeteroFuzz selectively invokes a hardware simulator only when necessary to reduce latency.

HeteroFuzz identifies and prioritizes kernel-sensitive inputs for mutation.

Code Coverage uses → 100% 2e

HeteroFuzz monitors code coverage as one of its guidance mechanisms in addition to accelerator spectra.

AFL ← compares with 100% 2e

HeteroFuzz is compared to AFL-like naive fuzzing throughout the evaluation.

HeteroFuzz evaluates and detects platform-dependent divergence across seven benchmark programs.

HeteroFuzz uses differential testing by comparing CPU vs FPGA execution results.

HeteroFuzz directly analyzes HLS synthesis pragmas to guide its fuzzing strategy.

HeteroFuzz monitors accelerator-relevant value spectra as a guidance mechanism.

HeteroFuzz designs and implements dynamic probabilistic mutations to increase the chance of hitting divergent behavior.

Amazon F1 mentions → 90% 1e

HeteroFuzz paper mentions Amazon F1 as a cloud provider supporting FPGA development frameworks.

Fuzz Testing implements → 100% 1e

HeteroFuzz is a fuzz testing tool implementing fuzz testing techniques for heterogeneous applications.

HeteroFuzz uses static backward slicing to identify kernel-sensitive inputs.

LLVM uses → 100% 1e

HeteroFuzz instruments the host program using LLVM based on extracted control flow graph.

control flow graph uses → 100% 1e

HeteroFuzz extracts and uses the control flow graph of the host program for coverage tracking.

Seed Input uses → 90% 1e

HeteroFuzz, like AFL, starts from seed inputs and mutates them to generate new test inputs.

Xilinx Vivado HLS uses → 100% 1e

HeteroFuzz experiments use Xilinx Vivado HLS to simulate kernel execution on FPGA.

The paper introduces HeteroFuzz as a novel fuzz testing tool for detecting platform-dependent divergence in heterogeneous applications.