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/

Zendy Plus

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

Zendy Tools

Zendy Open

Artificial neural networks (ANNs) have gained significant popularity in thelast decade for solving narrow AI problems in domains such as healthcare,transportation, and defense. As ANNs become more ubiquitous, it is imperativeto understand their associated safety, security, and privacy vulnerabilities.Recently, it has been shown that ANNs are susceptible to a number ofadversarial evasion attacks--inputs that cause the ANN to make high-confidencemisclassifications despite being almost indistinguishable from the data used totrain and test the network. This work explores to what degree finding theseexamples maybe aided by using side-channel information, specifically switchingpower consumption, of hardware implementations of ANNs. A black-box threatscenario is assumed, where an attacker has access to the ANN hardware's input,outputs, and topology, but the trained model parameters are unknown. Then, asurrogate model is trained to have similar functional (i.e. input-outputmapping) and switching power characteristics as the oracle (black-box) model.Our results indicate that the inclusion of power consumption data increases thefidelity of the model extraction by up to 30 percent based on a mean squareerror comparison of the oracle and surrogate weights. However, transferabilityof adversarial examples from the surrogate to the oracle model was notsignificantly affected.

Model Extraction and Adversarial Attacks on Neural Networks using Switching Power Information