Buffer overflow vulnerabilities in CUDA: a preliminary analysis

TL;DR

This paper investigates buffer overflow vulnerabilities in CUDA GPU software, demonstrating how attackers can corrupt data or control flow, raising security concerns amid increasing GPU software adoption.

Contribution

It provides the first preliminary analysis of buffer overflow issues in CUDA, highlighting potential security risks and attack vectors specific to GPU-accelerated applications.

Findings

Buffer overflows can corrupt sensitive data in CUDA programs.

Attackers can manipulate function pointers to control execution flow.

Security vulnerabilities in CUDA are a significant concern for widespread adoption.

Abstract

We present a preliminary study of buffer overflow vulnerabilities in CUDA software running on GPUs. We show how an attacker can overrun a buffer to corrupt sensitive data or steer the execution flow by overwriting function pointers, e.g., manipulating the virtual table of a C++ object. In view of a potential mass market diffusion of GPU accelerated software this may be a major concern.