Events

Hide Sidebar Show Sidebar
Hide Sidebar Show Sidebar

Current Seminar Series

September 13, 2025

No events scheduled

September 15, 2025

High Performance Extreme Computing Virtual Conference, Sept 15-19

Part Of

10:30A
- 7:30P

Location

TBD
Add to Calendar 2025-09-15 10:30:00 2025-09-15 19:30:00 America/New_York High Performance Extreme Computing Virtual Conference, Sept 15-19  The Boston Chapter of the IEEE Computer Society is joining with SIAM (the Society for Industrial and Applied Mathematics) as cooperating society co-sponsors of this year's High Performance Extreme Computing Virtual Conference September 15-19.  The primary organizer is MIT Lincoln Lab and its Supercomputing Center.The conference has over 150 speaker presentations (many involving CSAIL presenters or co-authors) and complementary (FREE) registration is available.  (There is also a regular registration available that includes conference proceedings for about $200, but for most people FREE is almost always a good deal.  See the registration page linked to below for details about differences in registration type.  Signing up for free registration is particularly worthwhile if you only expect to attend one or  a few presentations.) The conference will be held virtually on the Engagez platform, allowing for networking with other attendees across the globe.You can see an overview of the conference at <https://ieee-hpec.org/>.The registration page is at<https://ieee-hpec.org/index.php/conference-registration/>.The keynotes speakers includeDay 1 Keynote:  Dr. Nick Rotker (MITRE Chief BlueTech Strategist) --- Enabling Advances in OceanAIDay 2 Keynote:  Prof. Julie Shah (MIT AeroAstro Dept Head)Day 3 Keynote:  Dr. Ashley Conard (Microsoft) --- Building AI That Users TrustDay 4 Keynote:  Joshua Patterson (NVIDIA VP of Solutions Engineering) --- SPACE MICE: The Next  Generation Data SystemsDay 5 Keynote:  Prof. Bill Gropp (NCSA Director; AAAS, ACM, IEEE, NAE & SIAM Fellow) --- Performance Engineering with MPIOther CSAIL/MIT presenters include Chuchu Fan, Cathy Wu, Srini Devadas and Sara Beery.  Mike Stonebraker is a paper co-author and is on the conference advisory committee.Special sessions includeAge of Mixed-Precision: Algorithms, Libraries, and Applications; organizer: Dr. Piotr Luszczek (MIT LLSC & UTK ICL)Bridging Quantum and High Performance Computing; organizer: Prof. Devesh Tiwari (Northeastern Univ.)GenAI Opportunities and AI Challenges; organizer: Dr. Vijay Gadepally (MIT LLSC), Dr. Daniel Burrill (MIT LLSC), Dr. Christian Prothmann (MIT CSAIL)Fastcode @ HPEC; organizer: Dr. Bruce Hoppe (MIT)MIT/Amazon/IEEE Graph ChallengeGraphBLAS Forum to define standard building blocks for graph algorithms; organizers: Dr. Timothy Mattson (HLG), Dr. Ben Brock (Intel), and Dr. Scott McMillan (CMU SEI)BRAINS: Building Resilience through Artificial Intelligence for Networked Systems; organizers: Dr. Sandeep Pisharody (MIT LL) and Dr. Thomas Hardjono (MIT)Scaling Research Computing Education; organizers: Dr. Julie Mullen (MIT LLSC),  Lauren Milechin (MIT ORCD), and Dr. Hayden Jananthan (MIT LLSC)The full program is online at <https://ieee-hpec.org/index.php/ieee-hpec-2025-prelim-agenda>.  TBD

September 16, 2025

CSAIL Forum with Josh Tenenbaum

Joshua Tenenbaum
Brain & Cognitive Sciences & CSAIL, MIT

Part Of

12:00P
- 1:00P

Location

TBD
Add to Calendar 2025-09-16 12:00:00 2025-09-16 13:00:00 America/New_York CSAIL Forum with Josh Tenenbaum Please join us for the next CSAIL Forum, featuring Prof. Josh TenenbaumCSAIL Forum hosted by Daniela RusSpeaker: Joshua Tenenbaum, Professor of Computational Cognitive ScienceDate/time: Tuesday 12:00-1:00 EDT, September 16, 2025 Venue: Live stream via Zoom: Registration requiredBio: https://bcs.mit.edu/directory/joshua-b-tenenbaum  TBD

Beyond Hospital Walls: Transforming Future Healthcare using Sensing and AI

Dr. Lili Qiu
UT Austin & Microsoft Research Asia
2:30P
- 3:30P

Location

32-G449
Patil/Kiva Conference Room
Add to Calendar 2025-09-16 14:30:00 2025-09-16 15:30:00 America/New_York Beyond Hospital Walls: Transforming Future Healthcare using Sensing and AI Title: Beyond Hospital Walls: Transforming Future Healthcare using Sensing and AIAbstract: Healthcare does not begin nor end in hospitals. With the deep convergence of cutting-edge sensing technologies and artificial intelligence, medicine is shifting from passive treatment to proactive protection, from within hospitals to everyday life. In this talk, I will introduce several novel sensing modalities—including acoustic sensing and imaging, mmWave imaging, and non-invasive glucose monitoring—that enable continuous, seamless, and intelligent health monitoring. I will also share our recent progress in multi-modality AI for sensing, which integrates diverse signals to create richer insights. Unlike prior alignment approaches limited by scarce paired data, our model leverages partial pairings across modalities to overcome data scarcity. Together, these advances open new possibilities for early diagnosis and chronic disease management to make healthcare more proactive, personalized, and accessible beyond hospital walls.  Bio: Dr. Lili Qiu obtained her MS and PhD degrees in computer science from Cornell University. Her current research interests include wireless communication, wireless/wearable sensing, machine learning, systems, and healthcare. She worked at Microsoft Research Redmond as a researcher in the System & Networking Group from 2001-2004. She has been a Professor at the University of Texas at Austin in the Department of Computer Science since 2005. She has also been serving as an Assistant Managing Director at Microsoft Research Asia. Dr. Qiu is an IEEE Fellow, an ACM Fellow, a National Academy of Inventors (NAI) Fellow. She also served as the ACM SIGMOBILE chair, was named an ACM Distinguished Scientist and was a recipient of the NSF CAREER award, among a number of other honors. TBD

HCI Seminar - Ziv Epstein - Re-inventing the attention machine (& building the serendipity machine)

Ziv Epstein
MIT (Schwarzman College of Computing and School of Architecture)

Part Of

4:00P
- 5:00P

Location

32-D463
Add to Calendar 2025-09-16 16:00:00 2025-09-16 17:00:00 America/New_York HCI Seminar - Ziv Epstein - Re-inventing the attention machine (& building the serendipity machine) Abstract:Today, algorithmic systems such as social media feeds and generative AI systems increasingly mediate human interactions and experiences. But interactions with these black-boxes reflect the worst of us due to  impoverished objectives that amplify problematic content, induce algorithmic overreliance and monoculture. This funhouse mirror-room in turn raises new questions about representation, agency and creativity. Whose perspective and values are being implicitly and explicitly amplified by these algorithms? Where does accountability lie and insight come from? What do users really want in the long run and how do we encode that into machines? In this talk, I will discuss two lines of work. The first explores how to reinvent the engagement-based “attention machine” of social media the interfaces and algorithms by aligning them with users’ values. I will discuss the role of attention and distraction in browsing patterns online and how to design mitigations to fight misinformation at scale by shifting attention to accuracy. Then I will discuss how to measure which human values are being algorithmically amplified by social media algorithms, and if those align with people's own values.  The second explores the domain of generative AI for creative application, and how we can foster active and divergent interactions with generative models to foster “serendipity” by re-injecting randomness into the models. Together, this work underscores the promise of new forms of interactions with algorithmic systems that center human agency to produce prosocial outcomes.Bio:Ziv Epstein is a postdoctoral associate at MIT sitting between Schwarzman School of Computing (SCC) and History, Theory & Criticism (HTC) in the School of Architecture. His current research explores how to audit the values amplified by social media ranking algorithms, and how to steer these algorithms to align with human values. Beyond social media, he is also interested in the impacts of AI on creative production in settings such as visual media and interpretative labor (e.g. divination). He was previously a postdoctoral fellow at Stanford University (2023-2025), and received his PhD from the MIT Media Lab (2023) where his dissertation focused on new ways to operationalize and measure attention on social media and implications for fighting misinformation online. He has published papers in venues such as the general interest journals Nature, Science and PNAS, as well as top-tier computer science proceedings such as CHI and CSCW. His work has received widespread media attention in outlets like the New York Times, Scientific American, and Fast Company. He is also a practicing multimedia artist whose work has been featured in Ars Electronica, the MIT Museum, and Burning Man.This talk will also be streamed over Zoom: https://mit.zoom.us/j/97497384805. TBD

AI4Society Seminar - Diyi Yang - Designing AI for Society: Teaming, Teaching, Tailoring

Diyi Yang
Stanford University
4:00P
- 5:00P

Location

45-102
Add to Calendar 2025-09-16 16:00:00 2025-09-16 17:00:00 America/New_York AI4Society Seminar - Diyi Yang - Designing AI for Society: Teaming, Teaching, Tailoring Abstract: Recent advances in large language models (LLMs) have revolutionized how humans and AI systems work, learn and interact, creating new opportunities for collaboration while also raising new challenges. In this talk, we explore the evolving landscape of human–AI collaboration from three perspectives: teaming, teaching, and tailoring. The first part on teaming shows how collaboration matters by introducing Co-Gym, which supports and evaluates human–agent collaboration, and a national audit of worker preferences that highlights mismatches between what workers want and current technological capabilities. The second part on teaching presents CARE, a scalable training system that leverages large language models to upskill counselors through realistic roleplay and structured feedback. The tailoring part introduces GUM, a general user modeling architecture that infers and reasons about unstructured context from users’ computer use to enable proactive AI assistance. Overall, this talk highlights how to develop AI systems that are not just tools, but meaningful collaborators working alongside us, helping us grow, and adapting to who we are.Bio: Diyi Yang is an assistant professor in the Computer Science Department at Stanford University, also affiliated with the Stanford NLP Group, Stanford HCI Group and Stanford Human Centered AI Institute. Her research focuses on human-centered natural language processing and human-AI interaction.  She is a recipient of  Microsoft Research Faculty Fellowship (2021),  NSF CAREER Award (2022), an ONR Young Investigator Award (2023), and a Sloan Research Fellowship (2024).  Her work has received multiple paper awards or nominations at top NLP and HCI conferences. TBD

Sparsification of 1-in-3-SAT

Standa Živný
University of Oxford

Part Of

4:15P
- 5:15P

Location

32-G449
Refreshments at 4:00 PM
Add to Calendar 2025-09-16 16:15:00 2025-09-16 17:15:00 America/New_York Sparsification of 1-in-3-SAT I will introduce a new notion of sparsification that doesn't drop constraints but merges variables. Using tools from additive combinatorics, I will then show that 1-in-3-SAT admits a sub-quadratic sparsifier. As an application, I will present an improved approximation algorithm for finding a linearly-ordered colouring of 3-uniform hypergraphs. Based on joint work with B. Bedert, T.-V. Nakajima, and K. Okrasa, to appear in FOCS'25. TBD

September 17, 2025

Discovering Safe, Effective Drugs via Machine Learning and Simulation of 3D Structure

Ron Dror
Stanford University
11:30A
- 1:00P

Location

TBD
Projected in 32-G575
Add to Calendar 2025-09-17 11:30:00 2025-09-17 13:00:00 America/New_York Discovering Safe, Effective Drugs via Machine Learning and Simulation of 3D Structure Recent years have seen dramatic advances in both experimental determination and computational prediction of macromolecular structures. These structures hold great promise for the discovery of highly effective drugs with minimal side effects, but structure-based design of such drugs remains challenging. I will describe recent progress toward this goal, using both atomic-level molecular simulations and machine learning on three-dimensional structures.This talk is part of the MIT Bioinformatics Seminar Series. TBD

Optiver Recruiting & Quantitative Research Tech Talk

12:00P
- 1:00P

Location

32-G449
32-G449 (Kiva Seminar Room)
Add to Calendar 2025-09-17 12:00:00 2025-09-17 13:00:00 America/New_York Optiver Recruiting & Quantitative Research Tech Talk Please RSVP here ahead of the event: https://share.hsforms.com/1qKnAhgJ0ROWCtQdhvbcKLA4xtvs . Refreshments will be served so please register so we can have an accurate count for food. Join Optiver on September 17th in 32-G449 (Kiva Seminar Room at CSAIL) from 12-1pm for a technical talk on Quantitative Research within trading. What you’ll learn:How Optiver researchers and engineers build models to determine the fair value of traded assetsTechniques for combining multiple inputs to generate robust pricing modelHow pricing models are used to inform trading strategies and manage risk in a high-frequency environment Speaker: Souktik Roy, Quantitative Researcher, PhD in Mathematics from University of Illinois Urbana-Champaign TBD

Metric Embeddings with Outliers

Kristin Sheridan
UT Austin

Part Of

2:00P
- 3:00P

Location

32-G575
Add to Calendar 2025-09-17 14:00:00 2025-09-17 15:00:00 America/New_York Metric Embeddings with Outliers We study the design of embeddings into Euclidean space with outliers. Given a metric space (X, d) and an integer k, the goal is to embed all but k points in X (called the “outliers”) into ℓ2 with the smallest possible distortion c. Finding the optimal distortion c for a given outlier set size k, or alternately the smallest k for a given target distortion c are both NP-hard problems. We consider bi-criteria approximations. Our main result is a polynomial time algorithm that approximates the outlier set size to within an O(log^2 k) factor and the distortion to within a constant factor.We also show that the techniques we use in designing outlier embeddings into Euclidean space have the potential to extend to a wider variety of target embedding spaces. In particular, we consider probabilistic outlier embeddings, which involve probabilistic embeddings and only require that non-outlier pairs have low distortion in expectation over the randomness of the embedding. We show that for probabilistic outlier embeddings into hierarchically separated trees (HSTs), there is a polynomial time algorithm that takes in a finite metric space with a (k,c) probabilistic outlier embedding and produces a probabilistic outlier embedding with O(log^2 k) outliers and O(c) distortion. TBD

ML for drug discovery at Genesis Therapeutics

Christina Ji, Pranav Murugan, David Williams
Genesis Therapeutics
6:00P
- 7:00P

Location

1-190
Add to Calendar 2025-09-17 18:00:00 2025-09-17 19:00:00 America/New_York ML for drug discovery at Genesis Therapeutics Genesis Therapeutics is an industry-leading start-up in the ML-driven drug discovery space. At Genesis, we are integrating ML into many phases of the drug discovery process: from generating new molecules, to sampling protein-ligand conformations, to predicting properties such as potency and ADME. Genesis has built a state-of-the-art denoising diffusion model for protein-ligand structure prediction. Genesis has also developed ML models for molecular property prediction to accelerate the virtual screening process and de-risk drug discovery programs. Genesis Therapeutics has raised over $300M in funding from top technology and biotech investors. We are hiring for internship and full-time positions in ML research, software engineering, and computational chemistry at https://genesistherapeutics.ai/careers/ TBD

September 19, 2025

How to Verify Any (Reasonable) Distribution Property: Computationally Sound Argument Systems for Distributions

Tal Herman
UC Berkeley

Part Of

10:30A
- 12:00P

Location

32-G882
Add to Calendar 2025-09-19 10:30:00 2025-09-19 12:00:00 America/New_York How to Verify Any (Reasonable) Distribution Property: Computationally Sound Argument Systems for Distributions  As statistical analyses become increasingly central, there is a growing need to ensure their results are correct. Approximate correctness can be verified by replicating the entire analysis, but can we verify without replication? We focus on distribution testing problems: given samples from an unknown distribution, the goal is verifying that the distribution is close to having a claimed property. Our main contribution is an interactive protocol between a verifier and an untrusted prover who both have sampling access to the unknown distribution. Our protocol can be used to verify a very rich class of properties: the only requirement is that, given a full and explicit description of a distribution, it should be possible to approximate its distance from the property in polynomial time. For any such property, if the distribution is at statistical distance $\varepsilon$ from having the property, then the verifier rejects with high probability. This soundness property holds against any polynomial-time  strategy that a cheating prover might follow, assuming the existence of collision-resistant hash.For distributions over a domain of size $N$, the protocol consists of $4$ messages and the communication complexity and verifier runtime are roughly $\widetilde{O}\left(\sqrt{N} / \varepsilon^2 \right)$. The verifier's sample complexity is $\widetilde{O}\left(\sqrt{N} / \varepsilon^2 \right)$, and this is optimal up to $\polylog(N)$ factors (for any protocol, regardless of its communication complexity).Even for simple properties, approximately deciding whether an unknown distribution has the property can require quasi-linear sample complexity and running time. For any such property, our protocol provides a quadratic speedup over replicating the analysis.  TBD

Distribution Learning with Advice by Arnab Bhattacharyya

Part Of

2:00P
- 3:00P

Location

32-D463
Add to Calendar 2025-09-19 14:00:00 2025-09-19 15:00:00 America/New_York Distribution Learning with Advice by Arnab Bhattacharyya Abstract: We revisit the problem of distribution learning within the framework of learning-augmented algorithms. In this setting, we explore the scenario where a probability distribution is provided as potentially inaccurate advice on the true, unknown distribution. Our objective is to develop learning algorithms whose sample complexity decreases as the quality of the advice improves, thereby surpassing standard learning lower bounds when the advice is sufficiently accurate. Specifically, we demonstrate that this outcome is achievable for the problem of learning a multivariate Gaussian distribution N(μ, Σ) in the PAC learning setting. Classically, in the advice-free setting,  Θ~(d²/ε²) samples are sufficient and worst case necessary to learn d-dimensional Gaussians up to TV distance ε with constant probability. When we are additionally given a parameter T as advice, we show that O~(d²⁻ᵝ/ε²) samples suffices whenever  ‖√T⁻¹Σ√T⁻¹−I‖₁< ε d¹⁻ᵝ (where ‖ ‖₁ denotes the entrywise l1-norm) for any , yielding a polynomial improvement over the advice-free setting. We also show similar results for product distributions over the hypercube.Joint work with Davin Choo (Harvard), Philips George John (NUS), and Themis Gouleakis (NTU). TBD

September 22, 2025

RBC Talk/Recruiting: Foundation Model Challenges and Opportunities in Financial Services

12:00P
- 1:45P

Location

32-G449
https://luma.com/2fygnljp
Add to Calendar 2025-09-22 12:00:00 2025-09-22 13:45:00 America/New_York RBC Talk/Recruiting: Foundation Model Challenges and Opportunities in Financial Services CSAIL Alliances & FinTechAI@CSAIL Board Member Royal Bank of Canada (RBC) Borealis AI Group will be at CSAIL on 9/22 in Kiva to deliver a technical talk from Dr. Greg Mori as well as connect with interested students for job opportunities. Talk Title: Foundation Model Challenges and Opportunities in Financial ServicesMonday 9/22 in Kiva 32-G449 12-1pm EST. Food will be served so please register for accurate food order!Registration link for the event: https://lu.ma/2fygnljp Speaker: Dr. Greg Mori, VP, RBC AI Fellow at RBC Borealis and Adjunct Professor in the School of Computing Science at Simon Fraser University.Talk Abstract:  Financial services are at the core of our economy. Opportunities for machine learning abound in this space, from capital markets to insurance services to wealth management to lending to tools that assist clients in managing their money.  Modern machine learning methods have transformed industries, yet particular challenges exist in realizing the full potential of machine learning in financial services.  These include explainability, data imbalance, partial observations, distribution shift, and self-supervised learning in low-signal settings.  I will describe the ATOM foundation model, which specializes in learning from asynchronous event sequences, to maximally utilize the richness of transactional data in financial services.Speaker Bio: Greg Mori is VP, RBC AI Fellow at RBC Borealis, where he leads AI Research and Innovation.  He is also an Adjunct Professor in the School of Computing Science at Simon Fraser University.  He received a Ph.D. in Computer Science from UC Berkeley in 2004 and an Hon. B.Sc. in Computer Science and Mathematics from the University of Toronto in 1999.  He was a Visiting Scientist at Google in Mountain View, California in 2014-2015. He served as Director of the School of Computing Science at Simon Fraser University from 2015-2018.  Dr. Mori conducts research in computer vision and machine learning.  He received the ICCV Helmholtz Prize in 2017.  He was a Program Chair for CVPR 2020 and a General Chair for CVPR 2023.  At RBC Borealis his team builds AI-based products for financial services.  These include the award-winning NOMI Forecast and numerous other industry-leading machine learning solutions.Registration link for the event: https://lu.ma/2fygnljp  TBD

September 23, 2025

Explicit Lossless Vertex Expanders

Rachel Zhang
CSAIL, EECS

Part Of

4:15P
- 5:15P

Location

32-G449
Refreshments at 4:00 PM
Add to Calendar 2025-09-23 16:15:00 2025-09-23 17:15:00 America/New_York Explicit Lossless Vertex Expanders We give the first explicit construction of lossless vertex expanders. These are d-regular graphs where every small set S of vertices has (1-eps)d|S| distinct neighbors. Previously, the strongest known explicit vertex expanders were those given by Ramanujan graphs, whose spectral properties imply that every small set S of vertices has 0.5d|S| distinct neighbors.Based on joint work with Jun-Ting Hsieh, Ting-Chun Lin, Alex Lubotzky, Sidhanth Mohanty, Ryan O'Donnell, and Assaf Reiner. TBD

September 24, 2025

September 25, 2025

Learning, engineering, and targeting cell states in cancer

Ava Amini
Microsoft Research in Cambridge, MA

Part Of

7:00P
- 8:00P

Location

32-G449
MIT Room 32-G449 (Kiva)
Add to Calendar 2025-09-25 19:00:00 2025-09-25 20:00:00 America/New_York Learning, engineering, and targeting cell states in cancer Boston, Guatemala, Panama, and Peru Chapters of the IEEE Computer Society, Boston IEEE Engineering in Medicine and Biology Society (EMBS), New Jersey Coast and North Florida Sections of IEEE and GBC/ACM7:00 PM, Thursday, 25 September 2025Note Date Change from previous announcement!MIT Room 32-G449 (Kiva) and online via ZoomLearning, engineering, and targeting cell states in cancer            Ava AminiPlease register in advance for this seminar even if you plan to attend in person athttps://acm-org.zoom.us/webinar/register/WN_Msf8F_LXTcSD2mWpDeVx5AAfter registering, you will receive a confirmation email containing information about joining the webinar.Indicate on the registration form if you plan to attend in person.  This will help us determine whether the room is close to reaching capacity. We plan to serve light refreshments (probably pizza) before the talk starting at around 6:30 pm. Letting us know you will come inperson will help us determine how much pizza to order.We may make some auxiliary material such as slides and access to the recording available after the seminar to people who have registered.              Abstract:Cancer is often treated using a reductionist approach: distilled to an individual subtype, mutation, or phenotype. But fundamentally, cancers are complex ecosystems that necessitate systems-level understanding and intervention. Addressing this problem is equal parts biology and computer science. In Project Ex Vivo, a joint cancer research collaboration between Microsoft Research and the Broad Institute, we are envisioning a new, constructionist paradigm for precision oncology, one powered by the bottom-up integration of computation and experimentation to understand the complexity of cell state ecosystems in cancer. In this talk I will share our recent efforts to build AI models to better define, model, and therapeutically target cell states in cancer.Bio:Ava Amini is a Principal Researcher at Microsoft Research in Cambridge, MA. Her research focuses on developing new AI methods to understand and design biology, with the ultimate aim of realizing precision biomedicines that improve human health. She is a co-lead ofEx Vivo , a collaborative effort between Microsoft and the Broad Institute, that is focused on defining, engineering, and targeting cell states in cancer.In addition to research, Ava is passionate about AI education and outreach ??? she is a lead organizer and instructor for MIT Introduction to Deep Learning , an in-person and global course on the fundamentals of deep learning.Ava completed her PhD in Biophysics at Harvard University and the Massachusetts Institute of Technology (MIT), where she was advised by Sangeeta Bhatia at the Koch Institute for Integrative Cancer Research and supported by the NSF Graduate Research Fellowship. Ava received her Bachelor of Science in Computer Science and Molecular Biology from MIT.Directions to 32-G449 - MIT Stata Center, 32 Vassar Street, Cambridge, MA: Please use the main entrance to the Stata Center at 32 Vassar Street (the entrance closest to Main street) as those doors will be unlocked. Upon entering, proceed to the elevators which will be on theright after passing a large set of stairs and a MITAC kiosk. Take the elevator to the 4th floor and turn right, following the hall to an open area; 32-G449 will be on the left. Location of Stata on campus mapThis joint meeting of the Boston Chapter of the IEEE Computer Society and GBC/ACM will be hybrid (in person and online).Up-to-date information about this and other talks is available online at https://ewh.ieee.org/r1/boston/computer/. You can sign up to receive updated status information about this talk and informational emails about future talks at https://mailman.mit.edu/mailman/listinfo/ieee-cs, our self-administered mailing list. TBD

October 01, 2025

October 07, 2025

October 08, 2025

Approximating High-Dimensional Earth Mover’s Distance as Fast as Closest Pair

Lorenzo Beretta
IBM

Part Of

4:00P
- 5:00P

Location

32-G575
Add to Calendar 2025-10-08 16:00:00 2025-10-08 17:00:00 America/New_York Approximating High-Dimensional Earth Mover’s Distance as Fast as Closest Pair We give a reduction from $(1+\epsilon)$-approximate Earth Mover's Distance (EMD) to $(1+\epsilon)$-approximate Closest Pair. As a consequence, we improve the fastest known approximation algorithm for high-dimensional EMD. Here, given $p\in [1, 2]$ and two sets of $n$ points $X,Y \subset (\R^d,\ell_p)$, their EMD is the minimum cost of a perfect matching between $X$ and $Y$, where the cost of matching two vectors is their $\ell_p$ distance. Further, Closest Pair is the basic problem of finding a pair of points realizing $\min_{x \in X, y\in Y} ||x-y||_p$. TBD

October 14, 2025

October 20, 2025

LLMs unlock new paths to monetizing exploits

Edoardo Debenedetti
ETH Zurich

Part Of

12:00P
- 1:00P

Location

32-G882
Hewlett
Add to Calendar 2025-10-20 12:00:00 2025-10-20 13:00:00 America/New_York LLMs unlock new paths to monetizing exploits Abstract: We argue that Large language models (LLMs) will soon alter the economics of cyberattacks. Instead of attacking the most commonly used software and monetizing exploits by targeting the lowest common denominator among victims, LLMs enable adversaries to launch tailored attacks on a user-by-user basis. On the exploitation front, instead of human attackers manually searching for one difficult-to-identify bug in a product with millions of users, LLMs can find thousands of easy-to-identify bugs in products with thousands of users. And on the monetization front, instead of generic ransomware that always performs the same attack (encrypt all your data and request payment to decrypt), an LLM-driven ransomware attack could tailor the ransom demand based on the particular content of each exploited device.We show that these two attacks (and several others) are imminently practical using state-of-the-art LLMs. For example, we show that without any human intervention, an LLM finds highly sensitive personal information in the Enron email dataset (e.g., an executive having an affair with another employee) that could be used for blackmail. While some of our attacks are still too expensive to scale widely today, the incentives to implement these attacks will only increase as LLMs get cheaper. Thus, we argue that LLMs create a need for new defense-in-depth approaches.Bio: Edoardo Debenedetti is fourth year a PhD student in Computer Science at ETH Zurich, advised by Prof. Florian Tramèr. His research focuses on real-world machine learning security and privacy. Most recently, he's been looking into the security of AI agents, working on evaluation frameworks and defenses. He is currently a Research Scientist Intern at Meta and he recently worked as a Student Researcher at Google.Zoom info:   Meeting ID: 945 5603 5878   Password: 865039 TBD

October 23, 2025

Re-visiting Authorized Private Set Intersection: A New Privacy-Preserving Variant

Lilika Markatou
TU Delft
4:00P
- 5:00P

Location

32-G882
Add to Calendar 2025-10-23 16:00:00 2025-10-23 17:00:00 America/New_York Re-visiting Authorized Private Set Intersection: A New Privacy-Preserving Variant We revisit the problem of Authorized Private Set Intersection (APSI), which allows mutually untrusting parties to authorize their items using a trusted third-party judge before privately computing the intersection. We also initiate the study of Partial-APSI, a novel privacy-preserving generalization of APSI in which the client only reveals a subset of their items to a third-party semi-honest judge for authorization. Partial-APSI allows for partial verification of the set, preserving the privacy of the party whose items are being verified. Both APSI and Partial-APSI have a number of applications, including genome matching, ad conversion, and compliance with privacy policies such as the GDPR. TBD

October 28, 2025