Next generation wireless (NextG) communication radios operating on ultra-high frequency bands, such as millimeter-wave (mmWave) and terahertz (THz) spectrum, need to know the location of other radios in order to communicate with them. Beam sweeping methods, also known as initial access, are a popular solution but tend to be inefficient in how much time and spectrum resources they take away from fast networks. Worryingly, they could also be a security vulnerability. This project aims to study, understand, and model the fundamental vulnerability of beam sweeping methods to Denial of Service (DoS) attacks that introduce signal alterations intended to deny the accurate beam alignment between the transmitter-receiver pair. We will explore applying AI/ML methods to both understand the attacks and improve resilience. We will showcase our research using NextG Wireless Lab@Mason's mmW testbed and CCI xG Testbed at Virginia Tech.