David Evans

1/ Most membership inference attacks (MIAs) have seemingly converged to black-box settings, driven by empirical evidence and theoretical folklore suggesting black-box access was optimal. But what if this assumption missed something critical? 😨 tl;dr? It did 🧵

Temporally shifted data splits in membership inference can be misleading ⚠️ Be cautious when interpreting these benchmarks!

Wrote a blurb for my submission "adversarial rinsing" 😈🧼 to the 'Erasing the Invisible' competition organized by @bang_an_ @furongh & co - awesome job btw! 🎉 Ranked nowhere near the top lol but a great learning experience! anshumansuri.com/post/eti/

Best book to started is A Pragmatic Introduction to Secure Computation Despite working on MPC for almost 4 years, I refer to this pretty regularly and have my staff at both @hashcloak and @StoffelMPC read it. It's pretty accessible as long as you have high school to first 1 of uni math skills (calculus + linear algebra) securecomputation.org

Thanks, @JaydeepBorkar! I should probably update that someday - its more than 20 years old now, but mostly still true. Even harder now for new students to get an email noticed with all the GPT-generated emails around now, but can still make it clear you've put in some effort.

I’m recruiting PhD students interested in trustworthy ML and ML for security this cycle! Please reach out via email anytime or find me at @acm_ccs next week. Retweets are greatly appreciated!

Meet Assistant Professor Fnu Suya. His research interests include the application of machine learning techniques to security-critical applications and the vulnerabilities of machine learning models in the presence of adversaries, generally known as trustworthy machine learning.

🌟 Join Our First Lego League Teams! 🚀 🔹 Division 1: One spot open for Grades 5 and 6! 🔹 Division 2: Three spots open for Grades 7 and 8! Don't miss the chance to be part of an exciting journey! 🤖

Does white-box access help with membership inference? 🤔 tl;dr: Yes ✅ Come find Xiao (@xzhang_0427) at the High-dimensional Learning Dynamics (HiLD) Workshop at @icmlconf this Friday to learn more 📊🔍 Prior works (@alexsablay et al. proceedings.mlr.press/v97/sablayroll…) theoretically show how black-box access is optimal for membership inference, and these findings seem to align with folklore and empirical observations around MIAs. However, the underlying assumptions for this theoretical result do not hold for models trained with SGD. We theoretically show that parameter access helps. Our theory also prescribes an attack that can be used for privacy auditing without any reference models. Read the full paper: arxiv.org/pdf/2406.11544 #ICML2024

@Yuantest3 Congratulations Yuan!

I recently got tenure at UCLA! It is an incredible privilege to work with our outstanding students, collaborators, and colleagues. I am deeply grateful to my family, friends, mentors, and the supportive department and university!

@iamgroot42 @KhouryCollege @Northeastern @AlinaMOprea Thanks so much, Dr. Suri - and congratulations on a terrific dissertation and outstanding presentation! It has been an honor to be part of your work, and I am excited for all the things you will do at @KhouryCollege and beyond.

PhD-efinitely DONE! 🎓✨Thrilled to announce I've successfully defended my thesis on 'Inference Privacy in Machine Learning'! 🚀 Next stop: @KhouryCollege @Northeastern as a postdoc this fall, supervised by @AlinaMOprea. Excited for what's next. 🎉 Slides: tinyurl.com/anshuman-defen… I want to extend my heartfelt thanks to my advisor, @UdacityDave, whose mentorship, wisdom, and encouragement have been instrumental in my journey. I am deeply grateful for your unwavering support and guidance. Thank you to my committee members, @bigflywth @nandofioretto @congshen, and Giuseppe Ateniese, for the insightful questions and feedback. A big shoutout to my friends, collaborators, and everyone who believed in me. Your support has been incredible, and I couldn't have done it without you. To my family, your love and encouragement have been my foundation ❤️

@thegautamkamath @ccanonne_ @shortstein Congratulations! This is a much deserved award for highly impactful and beautiful work.

Excited to share that "The Discrete Gaussian for Differential Privacy," joint work with Clément Canonne (@ccanonne_) and Thomas Steinke (@shortstein), was selected for the 2024 Caspar Bowden PET Award! Honoured for this work to be recognized. 1/2

Our MIA evaluation research for LLMs has been accepted at @COLM_conf! 🥳 It's time to think beyond just exact-record membership for privacy audits. Hope to see you in Philly soon! Meanwhile, check out this fascinating collection of related works that have emerged recently 👇

We will be closed the week of the Fourth of July. Camps will start on July 8 at 8:30 am. Join us for coding and robotics camps for kids aged 6 to thirteen from July 8 to August 16. Register at: buff.ly/3UU8onG Happy Fourth of July! #SummerCamps #CodingRoboticCamps

@UdacityDave Thanks Dave, it was also an honor to be advised you and @Yuantest3 !

Congratulations Suya! Its been a privilege to advise you, and excited to see all the great things you will do at UTK!

The first two days of summer camp have been a smashing success! We still have spots available for next week, June 17 to June 21. Visit our website to sign-up buff.ly/3UU8onG . Hurry, spots are filling up quickly!

Excited to share our work on data minimization for ML! The principle of data minimization is a cornerstone of global data protection regulations, but how do we implement it in ML contexts? 🧵: Let's dive into some insights. 🔗: arxiv.org/abs/2405.19471

Dive into the universe of LEGO Spike Essential at our new location on 2171 Ivy Road, Charlottesville, Virginia. Join us from June 17 to 21, 8:30 AM - 1:00 PM or 5:00 PM, for children aged 6 and 7. Explore more at our website (buff.ly/457yAOJ) #LEGO #SpikeEssential

As an incoming Assistant Professor @UVA CS, I invite applications for multiple PhD and internship positions! Seeking passionate individuals for interdisciplinary research in Software Engineering, Formal Methods, and Machine Learning. For more details, visit wenxiwang.github.io/recruit.html. Reach out if you are interested, and/or spread the word!