Somesh Jha
@someshjha.bsky.social
I am a professor in the computer sciences at UW-Madison. My technical interests in trustworthy ML, formal methods, and security. My other interests are Indian classical music, mindfulness, tennis, and pickleball.
#TPDP2026 One of the co-chairs is Amrita who is alum of our group. Way to go Amrita.
Clément Canonne@ccanonne.github.io · 9mo agoICYMI: the 2026 Theory and Practice of Differential #Privacy Workshop (TPDP 2026) will be in Boston, June 1-2 2026. The deadline to submit a 4-page abstract is ⏰ February 18 AoE (note: the workshop is non-archival) More information about #TPDP2026: tpdp.journalprivacyconfidentiality.org/2026/
Last year SAGAI workshop had the most attendees out of all the workshops at IEEE S&P. Don't miss out.
Mihai Christodorescu@mihaichr.bsky.social · 10mo agoThe SAGAI workshop on secure agentic GenAI is returning in 2026 to the IEEE Symposium on Security & Privacy in San Francisco next May. New: the workshop will accept a limited number of research papers! Looking forward to your most exciting results. SAGAI’26 site: sites.google.com/view/sagai-2...
Happy Diwali to all. May this coming year be full of joy and prosperity.
a man is speaking into a microphone and says happy diwali .
ALT: a man is speaking into a microphone and says happy diwali .
media.tenor.com
Thanks for inviting me @simonsinstitute.bsky.social The audience interaction was incredible.
Simons Institute for the Theory of Computing@simonsinstitute.bsky.social · last yr.2/2 Somesh Jha (@someshjha.bsky.social) spoke of the need to view the safety of machine learning systems through the lens of security and cryptography at the Simons Institute's workshop on Theoretical Aspects of Trustworthy AI. Video (Part II): simons.berkeley.edu/talks/somesh...
LLM-integrated applications and agents are vulnerable to prompt injection attacks, in which adversaries embed malicious instructions within seemingly benign user inputs to manipulate the LLM’s intended behavior.
Recently, we received a DARPA grant on the problem of LLM-assisted translation of C to Rust. The team consists of amazing set of PIs from UW, Berkeley, UIUC, and Edinburgh. Really excited about what we can do. Full article can be found here: www.cs.wisc.edu/2025/07/15/t...
Translating Legacy Code for a Safer Future: DARPA Backs Effort to Convert C to Rust
The C programming language—powering everything from operating systems to aircraft control software—has long been a cornerstone of modern computing. But its flexibility comes at a cost: memory safety i...
cs.wisc.edu
I have interacted with @gautamkamath.com and highly recommend him for this position. Please vote for him.
Gautam Kamath@gautamkamath.com · last yr.ICML's election for their board of directors has begun. I've thrown my hat in the ring. Please consider voting for Gautam Kamath. I have experience with the governance of TMLR, COLT, and ALT, and I think I've demonstrated myself as a consciencious and engaged community member.
This research took a while to complete, but very proud of the result. Will do a detailed post soon.
ePrint Updates@eprint.ing.bot · last yr.Secure Noise Sampling for Differentially Private Collaborative Learning (Olive Franzese, Congyu Fang, Radhika Garg, Somesh Jha, Nicolas Papernot, Xiao Wang, Adam Dziedzic) ia.cr/2025/1025
SAGAI 2025 program is now complete. What an amazing program! Don't miss it. sites.google.com/corp/ucsd.ed...
SAGAI'25 @ IEEE S&P - Program
07:30 - 08:45 am
sites.google.com
Somesh Jha@someshjha.bsky.social · 2y agoSAGAI'25 will investigate the safety, security, and privacy of GenAI agents from a system design perspective. We are experimenting with a new "Dagstuhl" like seminar with invited speakers and discussion. Really excited about this workshop at IEEE Security and Privacy Symposium.
SAGAI'25 will investigate the safety, security, and privacy of GenAI agents from a system design perspective. We are experimenting with a new "Dagstuhl" like seminar with invited speakers and discussion. Really excited about this workshop at IEEE Security and Privacy Symposium.
SAGAI'25 @ IEEE S&P
Goal The workshop will investigate the safety, security, and privacy of GenAI agents from a system design perspective. We believe that this new category of important and critical system components req...
sites.google.com
Interesting! Didn't know that sifr and sunya are connected.
Peyman Milanfar@docmilanfar.bsky.social · 2y agoA cipher (or cypher) is an algorithm for performing encryption or decryption - the foundation of all of cryptography research and practice. Not widely appreciated is that the word cipher has its direct origin in the arabic word صفر (sifr) meaning zero. Andصفر itself roots to sanskrit शून्य (śūnya)
Eid Mubarak to anyone of my friends that celebrate it. www.youtube.com/watch?v=5hwX...
Mubarak Eid Mubarak | 4K Video | Tumko Na Bhool Payenge | Salman Khan, Sushmita Sen | Sonu Nigam
YouTube video by HoHo Entertainment
youtube.com
Excellent place to work!
Andreas Zeller@andreaszeller.bsky.social · 2y agoWant security in research? And research in security? Actually, why not have both? "In view of the current geopolitical landscape", #CISPA has opened a new round of faculty positions, including CySec, SE, ML, theory: jobs.cispa.saarland/jobs/detail/...
Lorenzo graduated from my group and did some cool work on system and network security during his Ph.D. Congrats, Lorenzo! Proud of you.
Lorenzo De Carli@ldklinux.bsky.social · 2y agoAn update: I have been approved for promotion to Associate Professor with tenure! It takes a village to raise a faculty though, and I could have not achieved this without the help of many. So I am grateful for the journey so far, both to the University of Calgary and all those who supported me.
Have you ever wondered: In federated learning, what if we could leverage clients' private data without compromising privacy—what more could we achieve? 🚀 We're excited to introduce SLVR (Securely Leveraging Client Validation for Robust Federated Learning). Paper: arxiv.org/pdf/2502.08055
arxiv.org
Nicholas Carlini moves to Anthrophic. nicholas.carlini.com/writing/2025...
Career Update: Google DeepMind -> Anthropic
TODO
nicholas.carlini.com
Have you wondered about the following question? Is it possible to design an image watermark that (a) preserves the robustness of deep watermarks and (b) meets a well-defined notion of unforgeability and public-detectability? We address this question in arxiv.org/pdf/2502.04901
arxiv.org
If you want me to respond to your email, please don't start the email with "Dear Sheldon":-)
Amazing program at @satml.org 2025 in now online.
Konrad Rieck 🌈@rieck.mlsec.org · 2y agoThe list of accepted papers for @satml.org 2025 is now online: 📃 satml.org/accepted-pap... If you’re intrigued by secure and trustworthy machine learning, join us April 9-11 in Copenhagen, Denmark 🇩🇰. Find more details here: 👉 satml.org/attend/
Finally, after few attempts our paper will appear in iacr communications of cryptology. cic.iacr.org/p/1/4/31 Our paper presents the first provable watermarking scheme for large language models (LLMS) with public detectability or verifiability
Publicly-Detectable Watermarking for Language Models
cic.iacr.org
I was saddened to learn about Prof. Bob Brayton's demise. Interesting anecdote. My advisor, Ed Clarke, had a grant with him (I think from SRC). After the events, he wanted to go to a resturant but insisted in going where students eat and not something fancy.
In-person PC meetings have become relics of the past. I remember them very fondly, and in my opinion in-person PC meetings led to better decisions. Thoughts?
SIGPLAN@sigplan.bsky.social · 2y agoBefore PLDI 2021, in-person PC meetings were the norm. The steering committee decided to reconsider, 4 years hence, whether to revive them. That bill has come due. We have survey results. blog.sigplan.org/2025/01/02/s...
Happy New Year to all! May 2025 be prosperous and full of great success.
Registration for @satml.org is open. The program is shaping up to be really interesting. eventsignup.ku.dk/ieeesatml202...
IEEE SaTML 2025
eventsignup.ku.dk
Technical question. Is there a notion of how "hard" a probability distribution is? Let me explain. Intuitively, a distribution induced by a probabilistic finite-state machine (FSM) is "harder" than one induced by a probabilistic CFG.
