While the emerging 6G networks are anticipated to meet the high-end service quality demands of the mobile edge users in terms of data rate and delay satisfaction, new attack surfaces and zero-day attacks continue to pose significant threats to their successful realization and rollouts. Traditionally, most service provisioning techniques considered security metrics separately from the Quality of Service (QoS) and Quality of Expectation (QoE) parameters. The QoS/QoE parameters include data throughput, experienced delay, tolerable latency, jitter, resource utilization rate, spectral efficiency, energy efficiency, fairness, and other emerging key performance indicators (KPIs). Also, there are various security attributes, such as encryption key strength, authentication strength, network anomaly score, privacy metric, and so on. Typically, the resource allocation optimization techniques to maximize the security aspects to protect the communication of mobile users or user equipment (UEs) have an adverse effect on the service quality. Therefore, a key research gap exists in balancing service quality and security levels in communication networks that has been either overlooked or identified in a rather scattered manner by researchers in the recent decade. Thus, a comprehensive survey of the state-of-the-art to clearly address this research gap and outline the possible solutions is yet to appear in the existing literature. In this paper, we address this by surveying the existing practices, challenges, and opportunities in the emerging 6G (i.e., beyond 5G) networks, where various AI (Artificial Intelligence)-based techniques such as deep learning meet the classical optimization techniques, to balance the service performance and security levels. Several networking topologies with relevant use-cases are included in the survey to discuss the existing and emerging trends of isolated as well as joint treatment of service and security levels. Lessons learned from each use-case are p...