Research & Development · July 2026

Cardano PubSub

In pursuit of a verifiable communication layer for the Cardano ecosystem — an innovation workstream at the frontier of peer-sampling and dissemination research.

Team

William WolffLead · Technical Architect

Denis FirsovSenior Formal Engineer

Ezequiel PostanSenior Prototyping Engineer

Jesus Diaz VicoApplied Cryptography Lead

Dana AlibrandiProduct Manager

Sandro Coretti-DraytonSenior Research Fellow

Introduction

About me

With IO for almost five years — first as Solution Architect, now as Technical Architect.
Worked on the backend architecture of the Lace wallet platform.
Led the Ouroboros Leios workstream through 2025 — now handed over to engineering.
Since early 2026: leading the PubSub workstream.

William Wolff

The workstream

What we're working towards

An innovation workstream: starting from research proposals and product-driven use cases, we design and prototype a Cardano-anchored publish/subscribe layer — identifiable publishers, verifiable events, on-chain registered topics, and an economic cost to creating publisher identities.

Node builders → node operators — authenticated emergency alerts from teams like IOG, Amaru, and the Cardano Foundation.
SPOs → delegators — announcements from topics that provably belong to the pool.
dReps → delegators — voting intent and rationale (CIP-1694).
dApps → users — critical alerts and dApp-specific, per-user notifications (e.g. open positions) under on-chain identities.

Trusted communication for Cardano

Before

Critical messages travel over X, Discord, Telegram — entirely off-chain.
Anyone can impersonate an SPO or a governance actor.
No canonical channels — audiences can't know where legitimate messages appear.
No delivery guarantee — a message may simply never reach its audience.
Spam and Sybil accounts cost the attacker nothing.

After

Publishers identified by on-chain credentials.
Every message cryptographically verifiable.
Topics registered on-chain — subscribers know what they subscribed to.
Delivery to every honest subscriber — failure risk formally analysed and bounded.
Fake publisher identities carry a real on-chain cost.

The guarantees Cardano gives transactions, extended to communication.

Phase 1 — exploratory research

From proposal to evidence

Where we started: the four use cases just shown, and the research paper SecureCyclon: Dependable Peer Sampling (July 2023).
The proposal: three layers — peer sampling, navigation, dissemination — security inherited across the layers.
What we did: tested those claims systematically — formal model checking, simulation, and manual analysis.
What we found: the sampled views drift from uniform, and a silent adversary can bias a node's view without a single protocol violation.
None of SecureCyclon's nine defences fire — the attack sits outside the paper's adversary model.
Where that led us: a simpler, modellable design — no peer sampling for now; on-chain registries close off the silent sampling attacks.

Analysed what we were given, widened to alternatives like Basalt — then narrowed to a modular base built to grow.

Phase 2 — the approach

Deriving the design, not proposing it

One bar to clear: a good graph — every message of every honest publisher reaches all other honest nodes.
Five candidate topologies (M1–M5), all facing the same adversary: a fraction μ of silent Byzantine nodes.
Two independent tracks: formal analysis — per-model laws for coverage, bandwidth, hops, and degrees, validated against Monte-Carlo simulation — and prototype experiments measuring the same models empirically.
Measured trade-off: good-graph probability (security) vs hops (latency) vs messages (bandwidth) — per configuration, not in the abstract.

If the final design carries complexity, it is because the data shows the simpler variants fail — and exactly why.

Model	Mechanism
M1	Push — every node, on first receipt, pushes the message to F random peers.
M2	Pull — each node pulls from RF forwarders, redrawn every epoch.
M3	Pull + publishing links — per-epoch pull plus s−1 publishing links carrying only the node's own publications.
M4	Bidirectional — each node picks RF peers uniformly; every pick becomes a symmetric link, and messages flood along all incident links except the one they arrived on.
M5	Directed — k_in inbound + k_out outbound links, both the node's own picks.

Direction

Next steps

Complete the formal analysis — validate the laws for each model and cross-compare them toward a recommendation.
Run the experiments — measure the five models in the prototype and analyze the results.
Design proposal by mid-August — a clear candidate we can put forward as the solution architecture.
If time permits, a draft CIP — an ambition, not yet a commitment.
Beyond this phase: economic feasibility — how the infrastructure pays for itself (fees, incentives) is an open question outside phase 2's scope.

Outcome: an evidence-backed architecture recommendation Cardano can adopt.

Cardano PubSub · July 2026

Thank you

Questions welcome.

Further information

pubsub.cardano-scaling.org github.com/input-output-hk/pubsub