Skip to content
STIMSMITH

CPU Bug Detection

Concept

CPU bug detection is the task of identifying defects in processor hardware before deployment. The provided evidence discusses it through RTL fuzzing, focusing on limitations of prior CPU RTL fuzzers and on INSTILLER, a proposed fuzzer that shortens instruction inputs via a variant of ant colony optimization (VACO), models multiple interruptions and exceptions more realistically, and uses hardware-based seed selection and mutation. Reported results include higher coverage, shorter inputs, more detected mismatches, and increased execution speed relative to prior work such as DiFuzzRTL.

First seen 5/24/2026
Last seen 7/11/2026
Evidence 6 chunks
Wiki v3

WIKI

CPU Bug Detection

CPU bug detection is the task of identifying defects in processor hardware before deployment. The provided evidence discusses this topic through the lens of RTL (register-transfer level) fuzzing, with emphasis on the practical limits of earlier CPU RTL fuzzers and on a proposed system called INSTILLER.

Context

READ FULL ARTICLE →

NEIGHBORHOOD

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

explore full graph →

RELATIONSHIPS

12 connections
INSTILLER ← evaluates 100% 6e
Instiller is evaluated on its ability to detect CPU bugs through RTL fuzzing experiments.
Spectre mentions → 95% 4e
CPU bug detection research mentions Spectre as a well-known example of a CPU hardware bug.
Meltdown mentions → 95% 4e
CPU bug detection research mentions Meltdown as a well-known example of a CPU hardware bug.
Ryzen Segfault Bug mentions → 95% 3e
CPU bug detection research mentions the Ryzen segfault bug as a reported hardware bug.
Pentium FDIV Bug mentions → 95% 3e
CPU bug detection research mentions the Pentium FDIV bug as a notorious hardware bug.
Broadwell MCE Bug mentions → 95% 3e
CPU bug detection research mentions the Broadwell MCE bug as a notable hardware bug.
INSTILLER ← uses 1e
Instiller is designed for CPU bug detection through RTL fuzzing.
Meltdown ← mentions 90% 1e
Meltdown is mentioned as a motivating example of why CPU bug detection is critical.
Spectre ← mentions 90% 1e
Spectre is mentioned as a motivating example of why CPU bug detection is critical.
Pentium FDIV Bug ← mentions 90% 1e
The Pentium FDIV bug is mentioned as a real-world CPU bug motivating detection work.
Broadwell MCE Bug ← mentions 90% 1e
The Broadwell MCE bug is mentioned as a real-world CPU bug motivating detection work.
Ryzen Segfault Bug ← mentions 90% 1e
The Ryzen segfault bug is mentioned as a real-world CPU bug motivating detection work.

CITATIONS

10 sources
10 citations — click to expand
[1] Bugs exist in hardware, such as CPU, and unlike software bugs these hardware bugs need to be detected before deployment. Instiller: Toward Efficient and Realistic RTL Fuzzing | SCIENCE@home
[2] Previous fuzzing work in CPU bug detection has the disadvantage that the length of RTL input instructions keeps growing, and longer inputs are ineffective for fuzzing. Instiller: Toward Efficient and Realistic RTL Fuzzing | SCIENCE@home
[3] INSTILLER is an RTL fuzzer based on ant colony optimization (ACO) that distills input instructions with a variant of ACO (VACO) to keep inputs short and efficient. Instiller: Toward Efficient and Realistic RTL Fuzzing | SCIENCE@home
[4] Related work cannot simulate realistic interruptions well in fuzzing, and INSTILLER solves the problem of inserting interruptions and exceptions in generating the inputs, handling multiple interruptions and exceptions with priorities. [2401.15967] Instiller: Towards Efficient and Realistic RTL Fuzzing
[5] INSTILLER proposes hardware-based seed selection and mutation strategies that use hardware-related heuristics and mutation operations to improve fuzzing performance in RTL fuzzing. [2401.15967] Instiller: Towards Efficient and Realistic RTL Fuzzing
[6] INSTILLER achieves 29.4% more coverage than DiFuzzRTL on real-world target CPU cores. Instiller: Toward Efficient and Realistic RTL Fuzzing | SCIENCE@home
[7] INSTILLER detects 17.0% more mismatches than prior RTL fuzzing work. [2401.15967] Instiller: Towards Efficient and Realistic RTL Fuzzing
[8] With the VACO algorithm, INSTILLER generates 79.3% shorter input instructions than DiFuzzRTL. Instiller: Toward Efficient and Realistic RTL Fuzzing | SCIENCE@home
[9] The input instruction distillation leads to a 6.7% increase in execution speed on average. [2401.15967] Instiller: Towards Efficient and Realistic RTL Fuzzing
[10] The paper implements a prototype named INSTILLER and conducts extensive experiments showing it outperforms previous work. [2401.15967] Instiller: Towards Efficient and Realistic RTL Fuzzing