A Deep Learning framework that analyses Windows PE files to detect malicious Softwares. the project includes:

• Sate of the art of the works done using machine learning or deep learning.
• The goals of this project in term of contribution to the researches
  • 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
• 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
  • RNN
  • CNN
  • FNN
• Discussing learning and test results
• Providing a document explaining all the steps and materials of this project