Published June 7, 2025
| Version v2
Conference proceeding
Open
SoK: Efficiency Robustness of Dynamic Deep Learning Systems
Authors/Creators
-
1.
The University of Texas at Dallas
Description
This artifact contains a comprehensive list of all efficiency attack papers (34 in total) included in the SoK, along with their key properties. It also provides implementations of three representative attack strategies, links to the datasets used, and code for executing the lightweight input transformation–based mitigation and detection techniques employed in the primary experiments. These resources support the findings presented in the USENIX Security ’25 SoK paper, “SoK: Efficiency Robustness of Dynamic Deep Learning Systems.”
Files
README.md
Files
(602.7 MB)
| Name | Size | Download all |
|---|---|---|
|
md5:6904e7b15fa130e5bbd0b7fdc356ad7f
|
16.5 kB | Preview Download |
|
md5:e664123204d9fc913e2bedc7df281fdc
|
602.6 MB | Preview Download |