McNdroid: A Longitudinal Multimodal Android Malware Detection Benchmark

McNdroid is a large-scale, longitudinal, multimodal dataset for Android malware detection designed to benchmark concept drift robustness. It spans samples collected from 2013 to 2025 and provides three complementary modalities: static feature vectors, API call graphs (GML), and JSON-based behavioral representations. The dataset was created by IQSeC-Lab.

Use Cases

• Benchmarking concept drift robustness in malware classifiers based on the longitudinal data from 2013-2025.
• Training multimodal detection models based on the described static, graph, and behavioral representations.
• Analyzing the evolution of Android malware features over time based on the 12-year sample span.
• Comparing the effectiveness of different feature modalities (static, graph, behavioral) for malware classification.

Strengths

• Longitudinal coverage spanning samples from 2013 to 2025.
• Multimodal design includes static feature vectors, API call graphs, and JSON-based behavioral representations.
• Includes a rich metadata CSV file alongside the core data modalities.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file formats, and dataset size are unknown, which may limit suitability assessment.
• License information is unknown, which may restrict usage.

Provenance

Source

IQSeC-Lab

Collection Method

Collection of Android malware samples over time.

Time Range

2013 to 2025

Freshness

Last updated 2026-04-30 07:56:09.