The purpose of this research is to assess the validity of using artificial intelligence to encrypt
and decrypt evasive malware capable of performing targeted attacks. This is done through the
creation and analysis of a prototype system of machine learning models. It has been found that
different machine learning models yield different results, but there are patterns and trends in the
abilities of all models. In the end, the system is successful in being able to use artificial intelligence
to encrypt and decrypt malware capable of accurately, but lacks in being a fully evasive malware
because of its inability to effectively hide trigger conditions. Although it has a drawback, there
are some considerations for future improvement to maintain the parts of the system that have been
proven to work, while effectively improving its shortcomings.
