- You find the web UI in the folder Deep4MalWeb
A Deep Learning framework that analyses Windows PE files to detect malicious Softwares. the project includes:
- Sate of the art of the work done using machine learning or deep learning.
- A new approach for detection:
- Enhancing detection rate and reducing False positive rate
- Proposing a technique to garantee the evolution of the model
- Proposing a defense technique against 'Adversarial Attacks'
- Defining and implementing a framework to extract PE files representation, this includes:
- PE headers
- PE strings
- Opcodes sequences
- Opcodes stats
- Bytes n-grams
- API Calls
- Building a training data set
- Over 120.000 malwares
- Over 30.000 benign software
- Defining and implementing a Deep Learning architecture to learn on the extracted data
- SAE: n-grams of bytes
- RNN: sequences of opcodes
- CNN: exe to bytes image
- FNN: a submodule to the SAE
