advertisement
advertisement
HANNOVER MESSE 2019, 01 - 05 April
Homepage>Conference program >Defending Deep Neural Networks

Defending Deep Neural Networks

Location & Language

Hall 2, Stand C07

Language

English

Event Details

Type of event

Lecture

Topic

Artificial Intelligence, Digital Factory, Industrie 4.0

Event Host

Missed the Event?

So far no information materials were provided for this event. Please contact the organizer directly for further information.

Description

What is presented? 

We present a new approach, called Randomly Perturbed Ensemble Neural Networks (RPENNs), to defend artificial neural networks against adversarial input. We compare our approach against common defense mechanisms. We also benchmark our approach against several attack methods.

What is special? 

Common defense mechanisms are tailored towards one particular attack scheme. As soon as an attacker changes the attack approach, a common defense mechanism breaks down. Our defense mechanism is robust against a range of attacks, which enables us to achieve a higher robustness against attacks in general.

What are the potentials? 

Potentially, our method hardens systems, that use deep learning, against malicious input.

Speaker

 Prof. Dr. Dirk Nowotka

Prof. Dr. Dirk Nowotka

Christian-Albrechts-Universität zu Kiel

Dr. Dirk Nowotka leads the Dependable Systems group of the Computer Science department at Kiel University, Germany. Prior to joining Kiel as a Heisenberg-Professor in 2011, he was a research scientist ...

Go to profile

Stored items

0

advertisement
Server communication error: Item could not be saved.