Targeted Fuzzing for Unsafe Rust Code: Leveraging Selective Instrumentation

TL;DR

This paper introduces a selective fuzzing approach for Rust that targets unsafe code segments, significantly improving vulnerability detection efficiency without additional computational costs.

Contribution

It presents a novel method leveraging selective code coverage feedback to focus fuzzing on unsafe Rust code, enhancing efficiency and detection speed.

Findings

Improved fuzzing efficiency on unsafe Rust code.

No additional computational overhead during fuzz testing.

Faster detection of vulnerabilities in unsafe code segments.

Abstract

Rust is a promising programming language that focuses on concurrency, usability, and security. It is used in production code by major industry players and got recommended by government bodies. Rust provides strong security guarantees achieved by design utilizing the concepts of ownership and borrowing. However, Rust allows programmers to write unsafe code which is not subject to the strict Rust security policy. Empirical studies show that security issues in practice always involve code written in unsafe Rust. In this paper, we present the first approach that utilizes selective code coverage feedback to focus the fuzzing efforts on unsafe Rust code. Our approach significantly improves the efficiency when fuzzing Rust programs and does not require additional computational resources while fuzz testing the target. To quantify the impact of partial code instrumentation, we implement our approach by extending the capabilities of the Rust compiler toolchain. We present an automated approach to detect unsafe and safe code components to decide which parts of the program a fuzzer should focus on when running a fuzzing campaign to find vulnerabilities in Rust programs. Our approach is fully compatible with existing fuzzing implementations and does not require complex manual work, thus retaining the existing high usability standard. Focusing on unsafe code, our implementation allows us to generate inputs that trigger more unsafe code locations with statistical significance and therefore is able to detect potential vulnerabilities in a shorter time span while imposing no performance overhead during fuzzing itself.