IT Track
Prof. Giuseppe F. Italiano
Title of the Talk : Security and Privacy in Large Language and Foundation Models: A Survey on GenAI Attacks
Prof. G. Narahari Sastry
Prof. Sastry is an SS Bhatnagar awardee. He comes with expertise in AI applications in basic and translational research in Chemistry, Drug Discovery, Material Science, and allied areas.
https://www.neist.res.in/Biodata.pdf
Prof. Jean-Marie Gorce,
Title of the Talk: Learning networks: how to combine protocols, distributed algorithms and machine learning ?
DC Track
Prof. Michel Raynal
Title of the Talk : The Linearizability Hierarchy: An Example-based Introduction.
Dr. Debasish Das
Title of the Talk : Designing High-Performance Distributed Systems In Cloud
Keynote Speaker
DC Track
Prof. Michel Raynal (https://team.inria.fr/wide/team/michel-raynal/)
Professor at IRISA, University of Rennes, France
Title of the Talk : The Linearizability Hierarchy: An Example-based Introduction.
This talk is neither a survey nor a research paper in the classical sense, but an example-based introduction to the linearizability hierarchy. Its aim is to explain it in an “as simple as possible” way. While linearizability is a consistency condition that addresses objects defined by a sequential specification, set-linearizability and interval-linearizability are consistency conditions that have been introduced to take into account objects defined by a concurrent specification. So, they naturally extend linearizability (and its fundamental composability property) from sequential to concurrent objects specification. The aim of the talk is not to present the theoretical foundations of set-linearizability and interval-linearizability, but to explain concurrency patterns allowed by concurrent specifications, and show how these consistency conditions report on them. This is done in a very simple way with the help of three objects that define a family of snapshot objects. In addition to the fact that it constitutes a pedagogical introduction to the topic, this talk has also a concurrency-related historical flavor.
Keywords:
Asynchrony · Concurrent object · Concurrent specification · Contention point · Contention interval · Crash failure · Interval linearizability · Linearizability hierarchy · Object specification · Modular programming · Read/write register · Set-linearizability · Read/write -based communcation, Simultaneity · Snapshot object · Time ubiquity
Michel Raynal is an Emeritus Professor of Informatics, IRISA, University of Rennes, France. He is an established authority in the domain of concurrent and distributed algorithms and systems. Author of numerous papers on this topic, Michel Raynal is a senior member of Institut Universitaire de France, and a member of Academia Europaea. He was the recipient of the 2015 Innovation in Distributed Computing Award (also known as SIROCCO Prize), recipient of the 2018 IEEE Outstanding Technical Achievement in Distributed Computing Award, and recipient of an Outstanding Career Award from the French chapter of ACM Sigops. He is also Distinguished Chair Professor on Distributed Algorithms at the Polytechnic University (PolyU) of Hong Kong.
Michel Raynal chaired the program committees of the major conferences on distributed computing. He was the recipient of several ”Best Paper” awards of major conferences (including ICDCS 1999, 2000 and 2001, SSS 2009 and 2011, Europar 2010, DISC 2010, PODC 2014). He has also written 13 books on fault- tolerant concurrent (shared memory and merssage-passing) distributed systems, among which the following trilogy published by Springer: Concurrent Programming: Algorithms: Principles and Foundations (515 pages, 2013), Distributed Algorithms for Message-passing Systems (510 pages, 2013), and Fault-Tolerant Message-Passing Distributed Systems: An Algorithmic Approach Springer (459 pages, 2018). His last book titled Concurrent Crash-prone Shared Memory Systems: a Few Theoretical Notions (115 pages) has been published in 2022. Michel Raynal is also the Series Editor of the Synthesis Lectures on Distributed Com- puting Theory published by Morgan & Claypool.
More information can be found at https://team.inria.fr/wide/team/michel-raynal/
Dr. Debasish Das
Worked in Yahoo, Amazon and Google as senior technical contributor for the last 15 years, building high tps, distributed systems using AWS and GCP.
Title of the Talk : Designing High-Performance Distributed Systems In Cloud
Building high performance distributed systems is complex and needs the application of rigour and time tested practices. In this lecture, we shall discuss the common pitfalls of building such systems, more so in the cloud, which can’t be overcome by applying anecdotes, hearsay and just the work experience. In the first part of the discussion we would focus on the principles and design patterns involved in building a large-scale high performance system and in the second part we would bring in the complexity of building it in the cloud (using commodity hardware) where failure of the individual components or network partition is a common occurrence. In this discussion, we won’t focus on a specific application, but we will explain the concepts using 1) a simple web application which depends on a datastore and needs synchronous response 2) an asynchronous (off-line) data processing job. We shall discuss the theoretical issues of Scalability, Data Consistency and Robustness in the first part and focus on the problems created by the Faulty Hardware and Network Latency in the next. We would also discuss the Security issues in the cloud. We shall introduce the CAP theorem, Split Brain problem, Distributed Caching, Leader Election and Consensus algorithms, Gossip protocol and Clock Synchronisation. The goal of this talk is to shift the failure discussions to the left, while the architecture of the system is defined rather than doing it as an afterthought while the system starts failing in the production environment.
IT Track
Prof. Giuseppe F. Italiano (http://clio.luiss.it/team_/giuseppe-francesco-italiano/)
After getting a Ph.D. in Computer Science at Columbia University, Giuseppe F. Italiano worked as a Research Staff Member at the IBM T.J. Watson Research Center in Yorktown Heights (New York). When he was 33 years old, he won a National competition for Full Professor and went back to Italy. Before joining LUISS University, he was professor of computer science at University of Salerno, University “Ca’ Foscari” of Venice, and University of Rome “Tor Vergata”.
Title of the Talk : Security and Privacy in Large Language and Foundation Models: A Survey on GenAI Attacks
As Large Language Models and Foundation Models continue to evolve and integrate into various aspects of our society, understanding their capabilities, limitations, and potential vulnerabilities becomes of the utmost importance. In this paper, we aim at exploring the security and privacy landscape surrounding those models in a systematic way.
Prof. Jean-Marie Gorce,
Head of Telecommunications department, INSA-Lyon, Inria.
https://perso.citi.insa-lyon.fr/jmgorce
Title of the Talk: Learning networks: how to combine protocols, distributed algorithms and machine learning ?
Jean-Marie Gorce (Senior Member, IEEE) is a Professor with INSA de Lyon . He received the M.Sc. and Ph.D. degrees in electrical engineering from the Institut National des Sciences Appliquées (INSA), Lyon, France, in 1993 and 1998, respectively. He was a Co-Founder of the Centre for Innovation, Telecommunications and Integration of Services (CITI Lab), in 2001. He was a Visiting Scholar with Princeton University, Princeton, NJ, USA, from 2013 to 2014. He has been the deputy director for Science at INRIA Grenoble Rhône-Alpes from 2017 to 2021, and director for science at Inria Lyon from 2021 to 2024.
He leads the research group Maracas, associated to INSA and Inria. His research focuses on multi-user communicating systems, with approaches combining information theory, coding, distributed algorithms, signal processing and machine learning, and has been supported by several French and European sponsored projects. He is currently involved in the European project INSTINCT on joint communication and sensing for 6G, and he leads the Inria-Nokia Challenge LearnNet working at the interesection of learning and networking. He has co-published more than 200 conference and journal articles in major venues.
While machine learning has revolutionized the digital world, both in its applications and in its foundations, the deep interactions between network protocol design and decentralized algorithms are still under-explored. In this presentation, we study the problem from two complementary angles: the use of machine learning for networks, and the optimization of networks for learning. We will discuss some recent results related to distributed and federated learning, as well as the use of machine learning algorithms to control and optimize wireless networks.