LAMDA: 1 Million Longitudinal Android Malware Samples (2013-2025)

LAMDA contains over 1 million feature-engineered Android APK samples collected by IQSeC-Lab for malware detection and concept drift analysis. The dataset spans a longitudinal period from 2013 to 2025, excluding 2015, providing a benchmark for temporal model evaluation.

Use Cases

• Training binary classification models to distinguish between benign and malicious Android APKs
• Analyzing concept drift by evaluating feature distribution shifts across the 2013-2025 timeline
• Benchmarking model performance degradation over time using longitudinal labels

Strengths

• Over 1 million samples
• 12-year temporal span (2013-2025)
• Feature-engineered for immediate machine learning application

Limitations

• Complete absence of data for the year 2015
• Potential label noise inherent in automated real-world malware collection

Provenance

Source

IQSeC-Lab

Collection Method

Collected from real-world sources and feature-engineered

Time Range

2013-2025 (excluding 2015)

Freshness

Last updated February 2026; includes data samples through 2025.

Geography

United States