Explore a graduate research presentation that uncovers vulnerabilities in Machine Learning models designed for DNS Over HTTPS (DoH) tunnel detection. Delve into the susceptibility of cutting-edge DoH tunnel detection models to black-box attacks, utilizing real-world input data generated by DoH tunnel tools. Discover specific vulnerable features that model developers should avoid, and learn how these findings can be applied to evade most Machine Learning-Based Network Intrusion Detection Systems. Gain insights into the immediate and practical implications of this research for cybersecurity professionals and ML model developers.