Thorsten Eisenhofer

A Representative Study on Human Detection of Artificially Generated Media Across Countries (n=3,002) covering audio, image, and text 🥳 Our #SP24 @IEEESSP paper is now available arxiv.org/pdf/2312.05976… Together with @joool_f @FranziskaPiaH @thorstenholz @jonasricker @_thrsten

That's a wrap! The 16th AISec workshop went great! Three outstanding keynote speakers and two insightful poster sessions. I'm proud of this community. Thanks to all contributors! Special thanks to @carmelatroncoso @securingAI @mariojfritz @florian_tramer @xinyun_chen_ @acm_ccs

Get rid of bad reviewers? 🤔 Today, @_thrsten presents our work @USENIXSecurity on misleading the automatic assignment of reviewers at large conferences (e.g. #CVPR or #NeurIPS). Our attack generates adversarial papers that select their own reviewers. @bifoldberlin @CASA_EXC 1/5

Congratulations Dr. @_thrsten

Systems for automatic paper-reviewer assignments (e.g., TPMS - mila.quebec/en/publication…) are used during the reviewing process of conferences such as @icmlconf. In arxiv.org/abs/2303.14443, we explore how authors might try to game these systems. We propose an attack that adapts...

Congratulations Dr. @doreenriepel!

Hackers are revealing that quadcopter operators aren’t nearly as hidden as they might think, but researchers have released a tool that can intercept and decode that signal. wired.trib.al/CIKDIDe

Nico Schiller from Ruhr-Universität Bochum is presenting his work on Drone Security and the Mysterious Case of DJI's DroneID at #NDSS23, talking about how attackers could eavesdrop on the drone's over-the-air data traffic!

⁦@dagstuhl⁩ seminar on Security of Machine Learning - having a great time!

I am happy to announce that our paper "Password-Authenticated Key Exchange from Group Actions" has been accepted to CRYPTO 2022 - joint work with Michel Abdalla, Thorsten Eisenhofer @_thrsten, Eike Kiltz @crypto_theory and Sabrina Kunzweiler

This is what DJI’s Drone-ID packets look like. every drone broadcasts this data all the time. we managed to receive + decode the packets over the air. no encryption. (we saw some confusion around that before) @merlinchlosta 1/n

Ein ganz praktisches Beispiel ist der "Alexa-Hacking-Demonstrator" – PhD-Student @_thrsten vom @CASA_EXC zeigt, wie #Sprachassistenten so manipuliert werden können, dass sie versteckte Botschaften wahrnehmen. @ruhrunibochum

#ACSAC2020 Voice Security pt 2 -- Pivoting from defense to attack, @leaschnherr presents a robust approach for over-the-air adversarial examples to attack automatic speech recognition. Fittingly, the system is dubbed "Imperio," an unforgiveable curse! ; )

Want to hear about detecting deep fakes using frequency analysis? Check out my talk at #icml2020 (11 am CEST). 🧾 Paper: arxiv.org/abs/2003.08685 ⚙️ Code: github.com/RUB-SysSec/GAN… Joint work with @_thrsten, @leaschnherr, Asja Fischer, @DorotheaKolossa and @thorstenholz.

Achievement unlocked: @zdf_logo, a news program for kids, featured our work on smart speakers and my daughters watched it 😍

We performed a study of accidental triggers, i.e., sounds that should not have triggered a voice assistant. You can find some results at unacceptable-privacy.github.io. Joint work with @m33x, @leaschnherr, @_thrsten, @DorotheaKolossa and Jan Wiele

very happy to announce our #ICML2020 paper on better detecting #fakeimages with the help of frequency transforms :) thx to all for the great collaboration 😃 @joool_f @leaschnherr @thorstenholz @_thrsten @asjafischer @CASA_EXC @casafant