Mobile Edge Computing (MEC) is a key technology for delivering low-latency services to mobile and edge devices, supporting applications like autonomous vehicles and smart cities. However, traditional hardware-based middleboxes limit flexibility and scalability, leading to the adoption of Network Function Virtualization (NFV). NFV enables the deployment of network functions as software, optimizing resource allocation and reducing costs. This study proposes a proactive failure prediction and migration strategy for MEC environments. Using a Long Short-Term Memory (LSTM) algorithm optimized by Super SAPSO (Simulated Annealing Particle Swarm Optimization), the model forecasts server failures with improved accuracy, reducing False Alarm Rates and improving Failure Detection Rates. For migration, the Improved Sparrow Search Algorithm (ISSA) is applied, factoring in CPU, memory, and server security thresholds to identify suitable migration servers for Virtual Network Functions (VNFs). ISSA’s fitness function balances migration cost and time, minimizing the impact of security constraints on migration efficiency. Results show that ISSA with a security check improves migration success ratios, especially at lower Service Function Chaining (SFC) arrival rates, while keeping migration times consistent across arrival rates of SFCs. This approach ensures high success rates, and reduced costs offering a resilient and efficient solution for dynamic MEC environments and laying the groundwork for adaptive, secure, and resource-efficient edge computing systems.