Traditional detection methods struggle with evolving malware, and single-source datasets are no longer sufficient. We introduce BCCC-Mal-NetMem-2025, a multi-source dataset combining memory and network data, enriched with a benign behavior profiler (BUEBP) and the VolMemLyzer-V2 analyzer for feature extraction. Built from 2,000 samples across eight malware families, it supports a multilayer detection system that consistently outperforms existing methods and identifies unknown threats.