ReFuzz
ToolFirst seen 7/4/2026
Last seen 9/10/2026
Evidence 19 chunks
NEIGHBORHOOD
No graph connections found for this entity yet. It may appear in future ingestion runs.
explore full graph →RELATIONSHIPS
27 connectionsReFuzz is evaluated on BOOMV4 as the primary processor-under-test.
ReFuzz is a processor fuzzing framework that implements hardware fuzzing techniques.
ReFuzz leverages tests from prior processors as seeds to guide fuzzing of a new processor-under-test.
ReFuzz implements contextual bandit algorithms to adaptively select and reuse prior-processor tests.
The paper introduces ReFuzz as its core contribution, a novel adaptive fuzzing framework.
ReFuzz constructs and uses a coverage matrix for its test minimization step.
ReFuzz is compared against Cascade in coverage and vulnerability detection evaluations.
ReFuzz employs differential testing by comparing PUT outputs against a golden-reference model.
ReFuzz mutates prior-processor tests to generate new tests for the processor-under-test.
ReFuzz uses a test minimizer to remove redundant PP-tests before training the CB model.
ReFuzz uses incremental coverage as the reward signal for the contextual bandit algorithm.
ReFuzz uses the exploration-exploitation tradeoff of contextual bandits to balance test reuse and exploration.
ReFuzz uses Synopsys VCS as the industry-standard simulator to generate coverage databases.
ReFuzz is integrated with TheHuzz as one of the baseline fuzzers for generating PP-tests and evaluation.
ReFuzz is integrated with Cascade as one of the baseline fuzzers for generating PP-tests and evaluation.
ReFuzz is evaluated on the RSD processor and detected vulnerabilities in it.
ReFuzz categorizes and uses vulnerability tests that trigger known vulnerabilities in prior processors.
ReFuzz is compared against TheHuzz in coverage and vulnerability detection evaluations.
ReFuzz uses contextual bandits to select and mutate effective tests.
ReFuzz uses integer programming to solve the test minimization optimization problem.
ReFuzz uses branch coverage as its primary coverage metric.
ReFuzz uses the epsilon-greedy algorithm within its CB model to encourage exploration.
ReFuzz implements its CB algorithms using the MABWiser Python library.
ReFuzz selects and mutates tests from historic processor designs.
ReFuzz uses Chipyard as the simulation environment for multiple RISC-V processors.
ReFuzz uses Spike as the golden-reference model for differential testing.
ReFuzz is inspired by MINTS for its test minimization approach.