license: apache-2.0
pretty_name: MalNet Signal
size_categories:
- 1M<n<10M
Malnet Signal
MalNet dataset where the raw binaries have been downloaded from Androzoo and processed into 1D signal representations instead of byteplot images as described in this work.
A full code base for generating and modelling the data can be found here.
Malware Signals
Malware signals are 1D representations of the bytecode of an executable which act as an alternative to byteplot images as input to machine learning models. These signals can be statically extracted from various formated (e.g. EXE, APK), and used to train a 1D CNN for malware classification. By using a 1D representation of the binaries, more information from the original binary is preserved and the addition of spurious spatial correlation is avoided, resulting in improved downstream model performance. A comparison of malware signals with byteplot images is shown below:
1D CNNs
The 1D signal representations can be used to train 1D CNNs. 2D CNN architectures developed for byteplot images can easily be adated to operate on the 1D signals by squaring the kernel size and stride convolution parameters, with the 1D equivalent models being found to outperform their 2D counterparts.
Comparison
1D CNNs achieve state-of-the-art performance on the MalNet dataset outperforming their 2D counterparts:
| Model | Binary | Type | Family | ||||||
|---|---|---|---|---|---|---|---|---|---|
| F1 Score | Precision | Recall | F1 Score | Precision | Recall | F1 Score | Precision | Recall | |
| ResNet1DV2-152D-SE | .874 | .907 | .846 | .503 | .643 | .453 | .507 | .580 | .480 |
| SHERLOCK | .854 | .920 | .810 | .497 | .628 | .447 | .491 | .568 | .461 |
| ResNet18 | .862 | .893 | .837 | .467 | .556 | .424 | .454 | .538 | .423 |
| ResNet50 | .854 | .907 | .814 | .479 | .566 | .441 | .468 | .541 | .443 |
| DenseNet121 | .864 | .900 | .834 | .471 | .558 | .428 | .461 | .529 | .438 |
| Densenet169 | .864 | .890 | .841 | .477 | .573 | .433 | .462 | .545 | .434 |
| MobileNetV2(x.5) | .857 | .894 | .827 | .460 | .547 | .424 | .451 | .528 | .423 |
| MobileNetV2(x1) | .854 | .889 | .825 | .452 | .527 | .419 | .438 | .532 | .405 |
Citation
BibTeX:
@misc{wilkie2025signalbasedmalwareclassificationusing,
title={Signal-Based Malware Classification Using 1D CNNs},
author={Jack Wilkie and Hanan Hindy and Ivan Andonovic and Christos Tachtatzis and Robert Atkinson},
year={2025},
eprint={2509.06548},
archivePrefix={arXiv},
primaryClass={cs.CR},
url={https://arxiv.org/abs/2509.06548},
}
APA:
Wilkie, J., Hindy, H., Andonovic, I., Tachtatzis, C., & Atkinson, R. (2025). Signal-Based Malware Classification Using 1D CNNs. arXiv [Cs.CR]. Retrieved from http://arxiv.org/abs/2509.06548