Previously undiscovered or unknown cyberthreats, known as zero-day attacks, are a serious threat to private organizations, governments, and individuals. The threat these zero-day attacks pose is so great due to the lack of knowledge about the threat and the lack of time to prepare for the threat. Conventional detection and prevention technologies such as signature-based detection and software patching both rely on prior knowledge and time to react to cyberthreats. Machine learning allows for cybersecurity professionals to not be dependent on these constraints. Eliminating the reliance on prior knowledge and response time is a crucial step in detecting, preventing, and combatting zero-day attacks. This work will discuss the threat these zero-day attacks pose to Internet security, explain the benefits of incorporating machine learning techniques into the detection, prevention and combatting of zero-day attacks, and will provide a recommendation of a comprehensive and industry wide defense-in-depth strategy to minimize the potential damage of zero-day attacks.