Erwin Quiring retweeted
Happy to share that our paper "AI-Generated Faces in the Real World: A Large-Scale Case Study of Twitter Profile Images" has been accepted at #RAID2024 🎉
Check out the preprint here: arxiv.org/pdf/2404.14244
Thanks to @Dede198904, @thorstenholz, Asja Fischer, and @ErwQui!
Just busy with celebrating ;) Honored to get the award @TECHTIDE_NI. Thanks to @mlsec and @tuBraunschweig for the great research fundament. Also nice to see that ML security---the topic of the thesis---is considered important.
I guess @ErwQui doesn't want you to know: He just received the "KI-Talente" award for his dissertation. The award recognizes the best AI-related PhD thesis in the state of Lower Saxony. @tuBraunschweig @CASA_EXC
Thrilled to announce that I've moved to @CASA_EXC @ruhrunibochum where I will continue my research at the intersection of ML and security. Special thanks to @mlsec for the excellent research time in the last years :-)
Erwin Quiring retweeted
We are really thrilled that our paper "Dos and Dont's of Machine Learning in Computer Security" received a Distinguished Paper Award at @USENIXSecurity 🏆
Joint work across @bifoldberlin @CASA_EXC @tuberlin @tuBraunschweig @KITinformatik @uclcs @kingscollegelon @IGSnews
1/3
Erwin Quiring retweeted
1/5 I will present a novel counterattack against deep-fake detection methods at the DLS workshop @IEEESSP today. Together with @mlsec @darpsec @ErwQui, we manipulated the frequency domain of deep fakes to evade detection. Link: arxiv.org/abs/2205.12543
Erwin Quiring retweeted
The program of DLS 2022 @IEEESSP is finally out!
@surrealyz and I can't wait to welcome you all on May 26 with a lineup of exciting research talks, including keynotes by @AlinaMOprea and @moyix and thought-provoking S&P panels (yes, S!)
Curious? Head to dls2022.ieee-security.org/
Erwin Quiring retweeted
We are excited to present a new event for our series on Machine Learning Security!
We will host Konrad Rieck (TU Braunschweig) on May 10th, 2022 at 17:00 CEST.
Registration and info here: eventbrite.com/e/machine-lea…
@mlsec
#adversarial #machinelearning #ai #security #mlsec
Erwin Quiring retweeted
Too much is happening on Feb 1 2022 AoE for the security community: USENIX Sec submission deadline, CCS R1 reviews, while trying to celebrate the Lunar New Year properly (best wishes 🥳), so...
...we extended DLS 2022 @IEEESSP deadline to Feb 8, 2022 AoE! dls2022.ieee-security.org
📢We've extended the @IEEESSP Deep Learning and Security workshop deadline to *Feb 8, 2022 AoE*!
One more week to prepare your awesome paper submissions! Come join us at the workshop! 🙂
dls2022.ieee-security.org/
Erwin Quiring retweeted
I have prepared a small tutorial with advices on "How to Review Literature in Systems Security", targeted at early career Ph.D. students, but also MSc/BSc students tackling their dissertations.
fabio.pierazzi.com/post/tuto…
I would be happy to hear also your advices and feedback 😉
Erwin Quiring retweeted
We are glad to announce the 1st edition of the ACM Workshop on Robust Malware Analysis (#WoRMA), co-located with #AsiaCCS 2022.
Deadline: January 12, 2022
More info: worma.gitlab.io/2022/
Erwin Quiring retweeted
This is an important #MLsec issue. If you don't understand how your ML system does what it does, it may not be doing what you think it is doing.
This quoted post is unavailable.
Knowing 15% of the defense is not enough: Our report about our #ML system which is based on diversification and got the first place in the defender challenge of the ML Security Evasion Competition is now online: arxiv.org/abs/2010.09569
The competition gave us the unique opportunity to examine evasion attacks / adversarial examples in a realistic black-box scenario with independent attackers. Thanks for organizing @msftsecresponse @drhyrum
5
Erwin Quiring retweeted
For the defender challenge:
(1) `domumpbq` from @mlsec's student team (Erwin, Lukas, Daniel) at @tuBraunschweig, who fended off 69% of evasion attempts!!
(2) `needforspeed` from @fabriciojoc and @MarcusBotacin at @SECRET_UFPR, who fended off 51% of attempts
Erwin Quiring retweeted
Adversarial EXEmples: two novel functionality-preserving attacks on (static) EXE malware detectors
Welcome RAMEn, a new light-weight formalization for crafting adversarial EXEmples, shipped with two novel powerful attacks: Extend and Shift!
Happy to share!
Paper: arxiv.org/abs/2008.07125
Code: github.com/zangobot/secml_ma…
@biggiobattista @DrScottCoull @zxgio @AlessanArmando Roli
This cute cat may fool your ML system!
Scaling attacks allow creating arbitrary outputs, with a huge impact on #ML systems. Join my talk @USENIXSecurity on Thursday for the first comprehensive analysis, incl. the root-cause and defenses. #usesec20 scaling-attacks.net
5/6 We identify the vulnerability underlying the attacks in theory as well as in practical implementations. This root-cause analysis allows us to understand why this attack is actually possible.
1
1






















