English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

As the number of parameters in Deep Neural Networks (DNNs) scales, the thirstfor training data also increases. To save costs, it has become common for usersand enterprises to delegate time-consuming data collection to third parties.Unfortunately, recent research has shown that this practice raises the risk ofDNNs being exposed to backdoor attacks. Specifically, an attacker canmaliciously control the behavior of a trained model by poisoning a smallportion of the training data. In this study, we focus on improving thepoisoning efficiency of backdoor attacks from the sample selection perspective.The existing attack methods construct such poisoned samples by randomlyselecting some clean data from the benign set and then embedding a trigger intothem. However, this random selection strategy ignores that each sample maycontribute differently to the backdoor injection, thereby reducing thepoisoning efficiency. To address the above problem, a new selection strategynamed Improved Filtering and Updating Strategy (FUS++) is proposed.Specifically, we adopt the forgetting events of the samples to indicate thecontribution of different poisoned samples and use the curvature of the losssurface to analyses the effectiveness of this phenomenon. Accordingly, wecombine forgetting events and curvature of different samples to conduct asimple yet efficient sample selection strategy. The experimental results onimage classification (CIFAR-10, CIFAR-100, ImageNet-10), text classification(AG News), audio classification (ESC-50), and age regression (Facial Age)consistently demonstrate the effectiveness of the proposed strategy: the attackperformance using FUS++ is significantly higher than that using randomselection for the same poisoning ratio.

Explore the Effect of Data Selection on Poison Efficiency in Backdoor Attacks