Dataset for: Predicting Original Malicious Code from Obfuscated Code Using Denoising AutoEncoders

This dataset accompanies the paper 'Predicting Original Malicious Code from Obfuscated Code Using Denoising AutoEncoders' (PeerJ Computer Science, under review). Contents: - benign.zip: RGB image representations of 10,000 benign software samples - Cleaned_Images.zip: DAE-reconstructed (denoised) RGB images - malicious.zip: RGB image representations of 7,000 malware samples All files are PNG images derived by mapping every 3 consecutive bytes of a binary executable to a single RGB pixel, reshaped into 128x128 images. No executable files are included. Contact: [e-mail]