Talk Title: The Cross-Chain Exchange Optimization Problem

Abstract: We propose a new distributed coordination problem: the Cross-Chain Exchange Optimization Problem. Consider a set of autonomous parties that own a collection of assets scattered across distinct blockchains. One party, the Solver, provides a recurring service where it proposes to the others a plan for redistributing their assets, claiming that such trades will not only leave each individual party better off, but that they will also maximize some collective measure of well-being (or minimize some collective penalty).

The Solver's proposal faces several challenges. If the proposal and the parties are honest, all parties should end up better off. But parties (including the Solver) are not trustworthy: in fact, all but one may be dishonest, willing to use any means available to twist the outcome to its own advantage. Can the Solver trick the other parties into an unfair exchange? If not, can the proposed exchanges be executed safely even if there are no built-in mechanisms for coordinating atomic actions across distinct chains?

This talk presents a novel, constant round complexity, cross-chain exchange optimization protocol. When an honest Solver devises a plan by solving a convex optimization problem, it also produces a certificate of optimality that can be efficiently verified. Using a combination of certificates, cryptographic signatures, and escrows, the parties execute a distributed protocol that succeeds if all parties are honest, and leaves no honest party worse off even if the protocol is corrupted by dishonest parties.

Bio: Maurice Herlihy has an A.B. in Mathematics from Harvard University, and a Ph.D. in Computer Science from M.I.T. He has served on the faculty of Carnegie Mellon University and the staff of DEC Cambridge Research Lab. He is the recipient of the 2003 Dijkstra Prize in Distributed Computing, the 2004 Gödel Prize in theoretical computer science, the 2008 ISCA influential paper award, the 2012 Edsger W. Dijkstra Prize, and the 2013 Wallace McDowell award. He received a 2012 Fulbright Distinguished Chair in the Natural Sciences and Engineering Lecturing Fellowship, and he is fellow of the ACM, a fellow of the National Academy of Inventors, the National Academy of Engineering, and the National Academy of Arts and Sciences. In 2022, he won his third Dijkstra Prize.Between 2022 and 2025, he was appointed Fellow-Ambassadeur of the Centre national de recherece scientifique (CNRS) in France. In 2025, he received honorary doctorates from the Università della Svizzera italiana, and from Technion (Israel Institute of Technology).

Talk Title: Randomization in Shared Memory Algorithms

Abstract: Designing and analyzing asynchronous shared memory algorithms is already difficult, but adding randomization adds a whole other layer of complexity to it. In this talk, we will discuss how our deterministic notions of progress and complexity become more nuanced, including the zoo of adversaries that randomization unleashes. The talk will aim to be as accessible as possible for Friday-after-a-conference brains!

Bio: Dante Bencivenga is a PhD candidate at the University of Calgary in Canada. He works on randomized shared memory algorithms under the supervision of Professor Philipp Woelfel, including co-authoring the PODC 2026 Best Paper "Simple and Efficient Randomized Wait-Free Locks". Dante also has a background in quantum algorithms and quantum cryptography.

Talk Title: Managing Memory in Concurrent Approximate Nearest Neighbor Search

Abstract: Safe memory management techniques determine when memory in a concurrent data structure may be freed or repurposed, given that other threads may still hold references to it. Existing protocols generally rely on the assumption that an object is retired only after it has been unlinked, so that any thread still holding a reference can be waited out or made to restart. Recently, we introduced CleanANN (SPAA’26), a graph-based approximate nearest neighbor (ANN) index supporting concurrent full dynamism that relaxes this assumption. Graph-based ANN indexes are generally directed graphs stored as adjacency lists. Data is associated with each node and edges enable efficient searches. Establishing that a logically deleted node is unreachable in order to retire it requires handling every incoming edge, introducing a pain point for supporting deletion. Our system instead accepts recycling a logically deleted graph node while other nodes still point at it, without a grace period or detection for concurrent threads. Safety comes from a per-node status which makes reader-triggered retirement well-defined, and we empirically studied the effectiveness in terms of query approximation quality for ANN. Combined with other techniques proposed, CleanANN achieves 1.14x-1.26x overall throughput compared to existing state-of-the-art. In this talk, we present these results and examine the key properties of ANNS that admit these optimizations.

Bio: “Sylvia” Ziyu Zhang is a PhD candidate at Massachusetts Institute of Technology advised by Prof. Julian Shun and Prof. Mike Cafarella. Her current research interests revolve around index design for high dimensional vector data and their application in modern data management systems. She has also worked on the theoretical aspect of algorithms for high dimensional data.

Talk Title: Information Leakage in Concurrent Objects

Abstract: An atomic object’s specification determines which return values are permitted in a correct execution. Its concurrent implementation, however, may reveal additional information through interactions between threads. Thus, correctness does not necessarily guarantee security or privacy.

In this talk, we investigate information leakage in concurrent objects from two complementary perspectives. We begin with history independence, which limits what an external observer can learn by inspecting memory. Previously studied in sequential systems, this notion presents new challenges under concurrency. We then introduce non-leaking concurrent objects, which limit what participating threads can learn from their own observations while following the prescribed protocol. We explore the possibilities and limitations of both notions and compare the guarantees they provide.

Bio: Noa Schiller is a PhD student in Computer Science at Tel Aviv University, advised by Rotem Oshman at Tel Aviv University and Hagit Attiya at the Technion. Her research focuses on concurrent data structures, with a particular interest in the information their implementations may reveal. She is a coauthor of "Non-Leaking Concurrent Objects", which received the DISC 2026 Best Paper Award.

Talk Title: On the (im)Possibility of Wait-free Oblivious RAM

Abstract: Oblivious RAM (ORAM) is a cryptographic primitive that hides the access pattern of a client accessing a storage server, i.e., the storage server does not know which entries the client accesses, nor whether it is reading or writing at those entries. ORAM schemes are a popular topic in the crypto & security community, with many papers published every year. Interestingly, most of these works consider a single client interacting with a server, while a small portion consider multiple clients accessing a parallel ORAM, but require that clients never access the same entry concurrently. Recently, we introduced MVP-ORAM (NDSS’26), the first wait-free ORAM, albeit with weakened security guarantees. In this talk, we present this construction, the fundamental challenges of building concurrent ORAMs and hypothesise the impossibility of implementing wait-free ORAMs with standard security guarantees.

Bio: Alysson Bessani is a Professor at the University of Lisbon Faculty of Sciences, Portugal, and a member of the LASIGE research unit. He received his Ph.D. in Electrical Engineering from UFSC (Brazil) in 2006, was a visiting professor at Carnegie Mellon University (2010), and a visiting researcher at Microsoft Research Cambridge (2014). Alysson coordinated/collaborated on 13 international projects and co-authored more than 120 peer-reviewed publications on dependability, security, Byzantine fault tolerance, and cloud storage. His work has been recognized with multiple awards, including two DSN test-of-time awards. He is also the original developer of the BFT-SMaRt replication library (https://github.com/bft-smart/library) and a co-founder of the Vawlt cloud storage startup (https://vawlt.io). More information about him can be found at https://www.di.fc.ul.pt/~bessani/.