Projects
AI-based Malware Analysis
In this project, after converting a malicious code file into an image, an artificial intelligence model was designed and trained to determine whether the given file was malicious or not, based on the image and classify the malware family. There are various systems in malicious code, and the images obtained by converting them differ in texture. We could design a CNN-based model that could detect these differences to demonstrate improved performance.
Facial Mask Detection
Due to the outbreak of the COVID-19 virus, it has become mandatory to wear masks in various public facilities. In the recent trend, the government has eased distancing or lifted laws related to wearing masks outdoors, but the need for wearing masks is still high. Various prior studies can be found, such as the correlation between mask wearing and the degree of corona infection, and the development of mask wearing recognition devices.
From the above research results and social trends, it can be inferred that wearing a mask will become an essential element in everyday life until the end of COVID-19. This study aims to build an artificial intelligence model that can detect whether a mask is worn by conducting transfer learning based on ResNet50. Thus, the given problem corresponds to a binary classification problem that classifies the data into two labels: non-masked and masked.
An accuracy of 95\% or more could be obtained, and the processing speed was relatively sufficient to be used in an actual field.
