Securing Pdfs: An Innovative LSTM Algorithm for Image-Based Malware Detection

The proposed system aims to enhance the current approach to combating PDF malware by addressing a key vulnerability in existing systems—specifically, the generation of evasive variants capable of bypassing machine learning based classifiers. Unlike the current system, the proposed solution leverages a hybrid algorithm and Variational Autoencoder (VAE) approach. Notably, it incorporates a pre-trained model to significantly reduce training time without compromising accuracy. This innovative combination of techniques presents an efficient and effective solution for image-based malware detection. In comparative testing, our proposed system outperforms the existing system, demonstrating superior accuracy and faster training times. By integrating hybrid algorithms and VAE, our approach provides an advanced defense against the evolving landscape of PDF based mal- ware threats. Keywords: Variational Autoencoder (VAE), PDF-based malware threats