The study delves into data poisoning attacks, focusing on LF and FP types. LF attack failed to deceive the server, while FP attack proved effective. The experiments were conducted on CIC and UNSW datasets related to computer networks. The results showed significant differences between benign and manipulated datasets. The LF attack was easily detectable, while the FP attack remained undetectable. Various percentages of poisoning were tested, with FP attacks proving difficult to detect due to high accuracy and ASR values.
To Another Language
from source content
arxiv.org
Key Insights Distilled From
by Ehsan Nowroo... at arxiv.org 03-06-2024
https://arxiv.org/pdf/2403.02983.pdfDeeper Inquiries