The advancement of quantum computing has introduced novel cybersecurity risks, vulnerabilities, and challenges. However, there remains a significant gap in the availability of comprehensive, large-scale datasets that capture quantum-native characteristics, such as labeled attacks, multiattack scenarios, hardware-noise modeling, and multitenant environments. This article describes the quantum attack dataset (QAD), a findable, accessible, interoperable, and reusable (FAIR) resource that systematically synthesizes large-scale quantum attack scenarios over a simulated multitenant quantum execution environment under controlled noise and attack models. QAD implements seven attack scenarios: swap injection, gate insertion, depolarizing noise, amplitude damping, measurement tampering, pulse-level attacks, and thermal relaxation, selected to reflect prevalent threats in noisy intermediate-scale quantum (NISQ)-era devices. The dataset is organized into two complementary views: a 77-feature set derived from circuit structure and measurement statistics, and generation metadata capturing attack parameters and hardware profiles, released separately to prevent label leakage. Two one-million-sample datasets are provided: a balanced configuration for method development, and an unbalanced configuration reflecting a literature-informed threat scenario distribution. Technical validation using four machine-learning baselines reveals that the discriminative challenge is concentrated in two attack families, swap injection (F1 ≈ 0.82–0.84) and pulse-level attacks (F1 ≈ 0.87–0.89), while five of the eight classes saturate at F1 = 1.00 across all models. The meaningful baseline is a noise-blind operating point (weighted F1 ≈ 0.88) that excludes a generation-time correlate; a full-feature upper bound (weighted F1 > 0.97) is also reported but should not be read as evidence of uniform task difficulty. Data files, schemas, and scripts are hosted in a persistent repository with detailed metadata, quality checks, and usage notes, enabling researchers to reproduce pipelines, extend attack scenarios, and apply QAD to future quantum-resistant security studies.
