IEEE Logo Itsoc Logo

Themed Sessions

ITW 2027 will host four themed sessions:

 

Post-Quantum Cryptography (PQC) from Information-Theoretic Perspectives: Lattice-Based and Code-Based Schemes

Organizer: Amin Sakzad (Monash University, Australia)

Abstract: Post-quantum cryptography (PQC) has moved from a theoretical research program to an urgent deploymentchallenge. The standardisation of post-quantum public-key primitives has accelerated the transition away from RSA and elliptic-curve cryptography, whose security would be threatened by a sufficiently scalable quantum computer. This makes ITW 2027 an ideal venue for a focused themed session on the mathematical, algorithmic, and information-theoretic foundations of PQC.

This proposed session will focus on two central families of post-quantum schemes: lattice-based cryptography and code-based cryptography. Lattice-based schemes, including module-lattice key encapsulation and signature mechanisms, offer strong efficiency and compactness properties, and are now at the center of near-term post-quantum deployment. Code-based cryptography, whose roots go back to the McEliece cryptosystem, provides a complementary security foundation based on the hardness of decoding problems. The coexistence of these two families is scientifically important: they rely on different hardness assumptions, have different failure probabilities and implementation tradeoffs, and raise different questions about decoding, noise, leakage, and reductions.

The topic is particularly well aligned with the IEEE Information Theory Workshop (ITW). Many of the central problems in PQC are naturally information-theoretic: decoding random and structured codes, estimating error probabilities in noisy algebraic channels, understanding secrecy and leakage under side information, AND quantifying decryption-failure probabilities, and designing reductions that connect computational hardness with statistical indistinguishability. Bringing together work on lattices, coding theory, cryptographic security proofs, and implementation-aware analysis will help clarify which information-theoretic tools can strengthen the next generation of post-quantum schemes.

 

Entropy, Combinatorics and Geometry

Organizer: Lampros Gavalakis (University of Cambridge)

Abstract: Entropy has become not only a useful tool, but also a central object of study in combinatorics and (convex) geometry, across both continuous and discrete settings. A prominent discrete example is the recent breakthrough proof by Gowers, Green, Manners, and Tao of Marton’s conjecture, also known as the polynomial Freiman–Ruzsa conjecture, in bounded-torsion abelian groups. The main idea of the approach is to prove a sufficient inequality for the Shannon entropy, and the key step in the proof is a data-processing argument which works for entropy but would not work for cardinalities, illustrating the effectiveness of information-theoretic approaches.

There has also been sustained interaction between entropy and convex geometry. The Entropy Power Inequality (EPI) may be viewed as an entropic counterpart of the Brunn–Minkowski inequality, and this analogy has led to forward and reverse entropy power inequalities, entropic formulations of volume estimates, and results concerning log-concave measures, sections and projections. More recently, the affirmative resolution of Bourgain’s slicing problem by Klartag and Lehec combined several convex-geometric and probabilistic ingredients—including stochastic localization and Milman’s theory of M -ellipsoids—with quantitative stability estimates for the EPI.

 

Coding for Next-Generation Data Storage

Organizers: Han Mao Kiah (Nanyang Technological University, Singapore) and Hengjia Wei (Xi’an Jiaotong University, China)

Abstract: The rapid growth of data-intensive applications has created new challenges for reliable, efficient, and scalable data storage. Beyond traditional magnetic and solid-state storage systems, emerging technologies such as DNA-based storage, distributed cloud storage, and large-scale archival systems require novel coding-theoretic techniques to address reliability, repair efficiency, synchronization errors, constrained channel models, and data recovery under increasingly complex operating conditions.

This themed session will bring together researchers working on coding theory for modern data storage systems. Topics of interest include, but are not limited to, coding for DNA storage, constrained and combinatorial coding, coding for distributed storage and repair, synchronization-error correction, and locally recoverable codes. The session will highlight recent advances and identify shared challenges across storage technologies.

 

Information Theory for Real-Time Decision-Making, Control, and Learning

Organizers: Photios A. Stavrou (EURECOM, France) and Matteo Zecchin (EURECOM, France)

Abstract: Next-generation intelligent systems must observe, communicate, learn, reason, and act in real time under tight latency and communication constraints, while meeting demanding reliability requirements. These systems range from networked robots, distributed sensing platforms, and industrial cyber-physical systems to emerging networks of agents built on AI foundation models. Despite their diversity, they share a common feature: heterogeneous agents must coordinate by exchanging observations, beliefs, plans, and uncertainty estimates over communication networks that are shared, noisy, time-varying, and capacity limited.

In such settings, communication becomes an integral part of the inference and decision-making process, and cannot be treated as an isolated data-transfer problem. Information transmitted at a given time changes what other agents know, believe, and decide; their resulting actions then alter the physical or informational environment, shaping future observations and the learning dynamics. Communication, inference, learning, and control are therefore coupled through the evolution of the system. The central questions are not only how much information must be exchanged, but also which information is relevant, when it should be communicated, and how it influences the system performance and its reliability.

Classical information theory provides fundamental limits for communication and powerful principles such as source-channel separation and asymptotic optimality. These principles may be insufficient when encoding, transmission, inference, and action must be performed causally, decisions must be made in real time, and communication delays are comparable to the dynamics of the underlying system. In such settings, the objective is not merely to reproduce data faithfully, but to communicate information that is timely and relevant to the task. This is especially important when the system must satisfy stringent reliability or safety requirements. Average reconstruction error or average decision loss may then be inadequate measures of performance, since rare but severe failures can dominate operational risk. This motivates information-theoretic formulations based on worst-case guarantees and risk-sensitive objectives, together with explicit models of epistemic and aleatoric uncertainty. The resulting questions concern what information must be communicated, when it must be delivered, and with what fidelity to ensure reliable system operation.

Recent advances in causal and nonanticipative rate-distortion theory, zero-delay and finite-blocklength coding, rate-distortion-perception, information-constrained control, decentralized decision-making, task-oriented communication, communication-efficient learning, and uncertainty quantification are beginning to reveal a common set of principles for real-time control of intelligent systems. The themed session will bring together researchers working across information theory, control, learning, and sequential decision-making to examine these connections, identify shared abstractions, and characterize fundamental limits of these systems under stringent latency, reliability, and resource constraints.