Toward Secure Multitenant Quantum Computing: Circuit Affinity, Crosstalk Patterns, and Grouping Strategies

TL;DR

This paper characterizes interference patterns in multitenant quantum computing across IBM hardware, identifying circuit types and structural similarities to inform secure, efficient job scheduling.

Contribution

It provides the first empirical analysis of crosstalk patterns and structural predictability in multitenant quantum processors, guiding hardware-aware scheduling strategies.

Findings

Aggressive circuits like Grover's show significant interference.

Sensitive circuits like QFT are highly susceptible to crosstalk.

Structural similarity varies within and across hardware revisions.

Abstract

Multitenancy increases throughput and reduces costs in cloud-based quantum computing, but concurrent job execution introduces security risks through inter-circuit crosstalk. We characterize the structural predictability of these interference patterns across seven IBM superconducting processors, spanning Heron (r1-r3) and Nighthawk (r1) architectures and five different circuit types. We evaluate pairwise interactions, by applying the Structural Similarity Index (SSIM) and a structural $t$-statistic to the concurrent execution of five foundational quantum circuits (QAOA, Grover's, QPE, QFT, and ZZFeatureMap), we quantify behavioral consistency across disparate hardware. Our results identify three types of circuits: universally aggressive, universally sensitive, and cotenant-dependent circuits. Aggressive circuits, such as Grover's Algorithm, exhibit a statistically significant interference pattern, yielding a $t$-statistic range of $[1.37,2.61]$ relative to the standalone baselines across all tested pairings. Conversely, sensitive circuits, such as the Quantum Fourier Transform, demonstrate a disproportionate susceptibility to multitenant execution, showing high deviations from single-tenant computational behavior. We demonstrate that crosstalk signatures are highly consistent within architectural revisions--with intra-revision similarity reaching $0.77$ (Hr3) and $0.68$ (Hr2)--while inter-revision similarity drops to $0.43$. Furthermore, we identify a ``topological decoupling" between Heavy-Hex and square lattice systems, where structural similarity falls to $0.01$ between Heron r1 and Nighthawk r1. These findings provide an empirical foundation for hardware-aware schedulers to strategically pair jobs, maximizing system utilization while preserving computational integrity.