A Hybrid Deep Learning Approach for Detecting Zero-Day Malware Attacks.

EasyChair Preprint 3177

Abstract

Begun in 1988, malware detection continues to be a challenging research topic in this epoch of technology. The exponential rise of IoT devices and its consumers has parallelly increased the number of security breaches in recent times, posing a major security concern. Research studies in malware detection analysis have proved that both dynamic and static analyses are time-consuming, inefficient and ineffective to detect novel malware signatures. The cybercriminals make use of evasive techniques like polymorphism and code obfuscation to alter the malware behavior rapidly and bypass malware detection. To countermeasure the cyber-attacks, machine learning algorithms (MLA’s) have come into the picture. The feature learning technique used by MLA’s to detect novel malware signatures turns out to be time-consuming. To bypass the feature engineering phase, we introduce the deep learning methodologies such as long short-term memory (LSTM) and convolutional neural networks (CNN). We made use of the binary malware datasets to train the algorithms, and once the malwares are detected they are classified and categorized into their respective malware families by means of deep image processing techniques. The results obtained in this paper showcases the Bright side of the deep learning architectures by outperforming the machine learning algorithms.

Keyphrases: Deep Learning., Image Processing., Machine Learning Algorithm, Machine Learning., Malware detection., cyber crime., deep image processing technique, deep learning methodology, malicious malware binary, malware binary, malware detection

@booklet{EasyChair:3177,
  author    = {Shaik Moin Sharukh},
  title     = {A Hybrid Deep Learning Approach for Detecting Zero-Day Malware Attacks.},
  howpublished = {EasyChair Preprint 3177},
  year      = {EasyChair, 2020}}