A Study of Overflow Vulnerabilities on GPUs

Abstract

GPU-accelerated computing gains rapidly-growing popularity in many areas such as scientific computing, database systems, and cloud environments. However, there are less investigations on the security implications of concurrently running GPU applications. In this paper, we explore security vulnerabilities of CUDA from multiple dimensions. In particular, we first present a study on GPU stack, and reveal that stack overflow of CUDA can affect the execution of other threads by manipulating different memory spaces. Then, we show that the heap of CUDA is organized in a way that allows threads from the same warp or different blocks or even kernels to overwrite each other's content, which indicates a high risk of corrupting data or steering the execution flow by overwriting function pointers. Furthermore, we verify that integer overflow and function pointer overflow in struct also can be exploited on GPUs. But other attacks against format string and exception handler seems not feasible due to the design choices of CUDA runtime and programming language features. Finally, we propose potential solutions of preventing the presented vulnerabilities for CUDA.

A Survey of Techniques for Improving Security of GPUs

“…Di et al [32] study buffer overflow vulnerabilities in CUDA-based GPUs. By causing a stack overflow, the address of a function pointer can be changed.…”

“…Clearly, the GPUs from all major vendors are vulnerable to attacks. [20,31,32] Denial of service [12,33] GPU components Global memory information leakage [16,17,23,29,[34][35][36], side-channel [8,26], covert channel [15], buffer overflow [20,32] Shared memory information-leakage [16,17,29,34,36], side-channel [25,26], covert channel [15], buffer overflow […”

“…In the absence of memory-access protection mechanisms, an adversary can read/write memory regions to either leak or destroy data. Miele [31] and Di et al [32] discuss strategies for launching buffer-overflow attacks, their impact and limitations. Erb et al [20] propose the use of canaries to mitigate buffer-overflow attacks.…”

Securing GPU via region-based bounds checking

“…By running a malicious gadget code on an altered control flow path, attackers can collect sensitive data or escalate the attacker's privilege to the system administrator. Recent work has shown that GPUs are also vulnerable to buffer overflow attacks [12,14,47]. For example, research by Miele [45] describes GPU attack scenarios in which the attacker manipulates the function pointer by using a buffer overflow and successfully forces the GPU kernel to execute the malicious function.…”

Efficient electro-magnetic analysis of a GPU bitsliced AES implementation

“…GPU executes program in a Single-Instruction, Multiple-Thread (SIMT) fashion, so it is well suited for cryptographic applications deployed in cloud computing environment to provide the Securityas-a-Service (SECaaS). Unfortunately, GPU-based applications are vulnerable to many known attacks as proposed in Di et al (2016); Naghibijouybari et al (2018); Jiang et al (2016). Among those published vulnerabilities *Correspondence: zhouyongbin@iie.ac.cn 1 State Key Laboratory of Information Security, Institute of Information Engineering, Chinese Academy of Sciences, Beijing, China 2 School of Cyber Security, University of Chinese Academy of Sciences, Beijing, China of GPUs, side-channel vulnerabilities are the most serious ones due to their non-invasiveness to target devices.…”