# XIR: A Framework for Interoperability across Cross-Chain Protocols Based on a Verifiable Intermediate Representation

Crypto.world

> Cross-chain protocols currently work point-to-point: an application on blockchain A can only reach blockchain C if an explicit A→C configuration exists. Using a dataset of 24,983,410 mainnet…

**arXiv**: 2609.20010

**Published**: 2026-09-17

**Authors**: Yushen Li, Linpeng Jia, Jiaying Feng, Ziliang Liao, Yi Sun

## AI summary

## Overview

**Research area:** Blockchain interoperability and cross-chain protocol security, specifically how to compose existing point-to-point cross-chain connections into multi-hop and cross-protocol paths without losing verification evidence.

**Technical level:** Advanced. The paper combines a large-scale empirical measurement of cross-chain activity with formal system modelling, adversary assumptions, security properties, and a Solidity prototype.

**Scope:** The paper measures connectivity across 286 blockchains from six cross-chain protocols, then proposes and evaluates XIR, a framework built on a verifiable intermediate representation (V-IR) that composes existing protocol connections into same-protocol and cross-protocol multi-hop paths.

## What This Paper Is About

Cross-chain protocols currently work point-to-point: an application on blockchain A can only reach blockchain C if an explicit A→C configuration exists. Using a dataset of 24,983,410 mainnet cross-chain events collected between 1 January 2025 and 1 November 2025 from Axelar, CCIP, Hyperlane, LayerZero, Relay, and Wormhole, the authors show that only 11,935 of the 81,510 possible ordered pairs among 286 active blockchains (14.64%) are directly connected. XIR addresses this by binding one application message to an ordered record of authenticated protocol deliveries, so that intermediate blockchains can hand a message from one connection (or one protocol) to another while the destination can still verify the complete history.

## Key Contributions

1. **A verifiable intermediate representation (V-IR).** XIR defines a record X = ⟨R, ctx, cert, [C₁, …, C_h]⟩ that binds the application record, its security context, a source certificate, and an ordered list of per-delivery receipts, together with stable root and message identifiers (`rid`, `mid`) derived through domain-separated hashes.
2. **Protocol composition with explicit guarantees.** XIR Gateways, XIR Adapters, an XIR Registry, and an XIR Router reuse configured connections for same-protocol and cross-protocol paths. Under four assumptions (A1 protocol-delivery soundness, A2 XIR-component soundness, A3 configuration authenticity, A4 cryptographic binding), the analysis establishes four properties: G1 source authorization, G2 ordered evidence integrity, G3 policy enforcement, and G4 at-most-once delivery.
3. **An empirical reachability study.** A normalized dataset of 24,983,410 distinct events covering 286 blockchains and 15,965 protocol-labeled directed edges quantifies direct reachability at 14.64% and the potential of composition at 96.86%.
4. **An open-source prototype and evaluation.** A Solidity implementation with a stateful Python orchestrator, integrated with Hyperlane and LayerZero, completing 40,000 local deliveries, 780 adversarial case-runs, seven successful public-testnet deliveries, and a separate campaign of 22,000 executions measuring verification-history and multi-hop cost.

## Main Findings

- **Direct connectivity is sparse:** The measured graph contains 286 active blockchains and 11,935 directly connected ordered pairs out of 286 × 285 = 81,510 possible pairs, a direct reachability of 14.64%. Full direct connectivity would require 81,510 point-to-point configurations.
- **Same-protocol composition:** Composing connections within a single protocol reaches 46,187 ordered pairs, or 56.66% — adding 34,252 pairs beyond direct connections.
- **Cross-protocol composition:** Allowing protocol switches reaches 78,953 pairs, or 96.86% of the 81,510 maximum, adding a further 32,766 pairs.
- **Per-protocol closure:** Hyperlane has the largest same-protocol closure at 27,059 pairs (33.20%).
- **Configuration savings:** Reproducing the 78,953-pair target with direct connections would require 78,953 point-to-point protocol configurations, including 67,018 for pairs not directly connected. XIR avoids those 67,018 configurations, equivalent to 84.88% of the total required by a point-to-point baseline serving the same reachable pairs.
- **Architecture cost units:** XIR needs 286 XIR Gateway placements (one per blockchain, O(N)) and 493 blockchain–protocol integrations (one per protocol available on each blockchain), rather than O(N²) configurations. The paper reports these units separately and states that deployment cost is compared in gas in Section 5.2; the specific gas figures are not included in the available content.
- **Hop distribution:** Among the 78,953 reachable ordered pairs, shortest-path hops are 11,935 at one hop (15.12%), 59,711 at two hops (75.63%), 7,186 at three hops (9.102%), and 121 at four hops (0.1533%), with cumulative percentages of 15.12%, 90.75%, 99.85%, and 100.0%. Nearest-rank P50, P95, and P99 are two, three, and three hops; the directed diameter is four hops, attained by 121 pairs, and 2,557 ordered pairs remain unreachable.
- **Longer same-protocol paths exist:** CCIP has nine reachable pairs whose shortest same-protocol path is five hops.
- **Sensitivity to observation window and activity:** Monthly protocol-switching reachability ranges from 46.4% to 65.7%, the final rolling-90-day union reaches 74.2%, and the cumulative reference is 96.9%. Raising the minimum activity per directed protocol edge from one to 1,000 distinct events reduces the cumulative protocol-switching result from 78,953 to 18,913 pairs.
- **Dataset limitations are stated:** Coverage is incomplete because APIs limit queries, no query logs were kept, the CCIP feed is particularly sparse, and a row records a source-side event without confirming destination execution; an absent edge does not establish that a connection was unavailable.

## Methodology in Plain English

The authors first built an empirical picture of who talks to whom. They pulled event feeds from the explorers and APIs of six cross-chain protocols (AxelarScan, the Atlas CCIP feed, Hyperlane Explorer, LayerZero Scan, Relay, and WormholeScan), normalized chain aliases into canonical network identities, kept only mainnet events between different blockchains, and deduplicated by protocol and event identifier. The result is a directed, protocol-labeled multigraph in which vertices are blockchains and edges are observed direct connections; the release contains 60 protocol/month Parquet files with 18 fields plus an edge table and identity registry.

On that graph they defined three reachability relations: direct connections, the transitive closure within each protocol (same-protocol composition), and the transitive closure over all protocols combined (cross-protocol composition). Comparing the size of each relation to 286 × 285 gives the reachability ratios. A hop is one directed protocol connection, while a protocol switch at an intermediate chain is counted separately and does not add a hop.

For the system design, the authors split the work into four roles: XIR Gateways authenticate the application at the source, verify each incoming delivery at intermediate chains, and verify the full history at the destination; XIR Adapters translate between XIR and a specific protocol's message format and verifier interface; the XIR Registry stores endpoints, profiles, adapters, validity windows, and approved incoming adapters; the XIR Router selects an admissible path. Each accepted delivery produces a receipt that is appended to the ordered list and folded into a running prefix commitment, so the final protocol's callback is tied to the same operation and to all earlier verification results. They then proved that, under the four stated assumptions, destination acceptance implies the four security properties, and built a prototype in Solidity with a Python orchestrator to test it locally, on public testnets, and under adversarial case-runs.

## Why This Matters

**Impact on research.** The paper reframes cross-chain interoperability from "add more point-to-point links" to "compose the links that already exist." Its distinctive move is treating verification history itself as a first-class object that can be carried across protocol boundaries, which sits between single-protocol multi-hop work (IBC multi-hop channels, xRoute) and single-boundary verification strengthening (zkBridge, CAVer).

**Real-world applications:**

- **Cross-chain applications (OApps, Hyperlane routers) reaching long-tail chains** without registering a new trusted peer or remote domain for every destination.
- **Asset and message bridging between ecosystems that share no common protocol**, such as a Hyperlane-connected chain reaching a LayerZero-connected chain through an intermediate chain.
- **Treasury and payment routing**, where a stablecoin issuer wants to move value between two chains where only indirect paths exist.
- **Protocol-agnostic middleware and wallet infrastructure** that can select paths subject to a per-user security policy rather than a single protocol's trust assumptions.

**Industry relevance.** The 67,018 avoided configurations and the drop from O(N²) to O(N) XIR Gateway deployments plus 493 integrations give a concrete operational argument: connecting 286 chains directly is a quadratic configuration burden, while composition is linear in chains for a fixed protocol set. The empirical 14.64% versus 96.86% gap makes the business case legible to bridge operators and interoperability consortia.

## Future Directions

- **Reachability under stricter activity thresholds.** With a minimum of 1,000 distinct events per edge, cross-protocol reachability falls to 18,913 pairs, so the practical limit of composition over genuinely active connections remains open.
- **Economic and gas cost of multi-hop paths.** The paper separates configuration, gateway, and integration counts and points to a gas comparison, but the trade-off between hop count (P50 of two, diameter of four) and end-to-end cost needs fuller characterization.
- **Trust aggregation across heterogeneous verifiers.** XIR preserves each protocol's own verifier, but a path that switches protocols inherits the weakest link along it; how security levels assigned in the XIR Registry should compose over multi-hop paths is a natural next question.
- **Extending beyond the two integrated protocols.** The prototype covers only Hyperlane and LayerZero, leaving adapter design and evaluation for Axelar, CCIP, Relay, and Wormhole as future work.

## Target Audience

Researchers and engineers working on blockchain interoperability, bridge security, and cross-chain protocol design; protocol architects at bridge operators and interoperability consortia; and graduate students in distributed systems or applied cryptography who want a measurement-grounded example of composing heterogeneous trust systems. Readers need comfort with directed graphs, hash commitments, and cross-chain message-passing terminology.

## Authors’ abstract

Cross-chain protocols enable applications to exchange messages across blockchains. Under point-to-point configurations, communication depends on a direct connection between the source and destination blockchains, limiting blockchain reachability and requiring additional configurations to connect more blockchains. To quantify this problem, this paper analyzes approximately 25 million mainnet cross-chain transaction events collected from six protocols (Axelar, CCIP, Hyperlane,…

[The full abstract on arXiv](https://arxiv.org/abs/2609.20010)

[Read the original paper](https://arxiv.org/abs/2609.20010)

---

Canonical: https://crypto.world/paper/xir-a-framework-for-interoperability-across-cross-chain-protocols-based-2609.20010
