Revealing GPUs Vulnerabilities by Combining Register-Transfer and Software-Level Fault Injection

Data used in the 51st Annual IEEE/IFIP International Conference on Dependable Systems and Networks (DSN) 2021.

Authors

• Fernando Fernandes dos Santos - Universidade Federal do Rio Grande do Sul (UFRGS)
• Josie E. Rodriguez Condia - Politecnico di Torino
• Luigi Carro - UFRGS
• Matteo Sonza Reorda - Politecnico di Torino
• Paolo Rech - Politecnico di Torino

About the data

This repository contains almost all the data presented in the paper. In the following section, there is a link to each Jupyter notebook with the data.

Data used in the paper

The data is divided in the following categories:

• FlexGripPlus microbenchmarks error distribution
• Profiling and Architectural/Program Vulnerability Factor

Used frameworks and codes

We used in this paper the following frameworks:

• Original NVIDIA NvbitFI
• Modified NVIDIA NvbitFI
• FlexGripPlus

The codes used in this paper are from:

• Rodinia benchmark suite
• CUDA Samples
• Darknet

Citing

@INPROCEEDINGS{dsn2021,
    author={Fernando Fernandes dos Santos and 
            Josie E. Rodriguez Condia and 
            Luigi Carro and 
            Matteo Sonza Reorda and
            Paolo Rech},
    title={51st Annual IEEE/IFIP International Conference on Dependable Systems and Networks (DSN)},
    year={2021},
    month={June}
}