The Open Radio Access Network (O-RAN) architecture promises a programmable and intelligent mobile network, yet its openness introduces new vulnerabilities to sophisticated attacks that can compromise network performance. Traditional static security measures are ill-equipped to handle dynamic threats like resource-consumption attacks, which are prevalent in 5G environments. This paper addresses this challenge by proposing a closed-loop security control mechanism, implemented as a near-real-time xApp on the O-RAN’s RAN Intelligent Controller (RIC). We uniquely leverage network slicing not merely for service differentiation, but as a dynamic enforcement tool to automatically isolate malicious User Equipment (UE). The proposed Secure Slicing xApp monitors UE performance metrics, detects anomalous resource consumption, and triggers a re-slicing procedure to move the malicious UE to a resource-deprived isolation slice. We implement and validate this solution on an end-to-end, open-source O-RAN testbed. Experimental results demonstrate that our system can successfully detect and neutralize a resource-consumption attack in under one second, thereby protecting network resources and ensuring Quality of Service for legitimate users. Furthermore, our analysis reveals an optimal trade-off between the RAN reporting frequency and the system’s reaction time, providing crucial insights for practical deployment. This work validates the potential of the O-RAN architecture for building intelligent, automated security defenses and provides a practical framework for enhancing the resilience of next-generation networks.