The rapid transition of 5G and beyond-5G (5G/B5G) networks toward Service-Based Architecture (SBA), Software-Defined Networking (SDN), and Network Function Virtualization (NFV) has increased the vulnerability of core-network components, particularly the User Plane Function (UPF), to Distributed Denial-of-Service (DDoS) attacks. Existing detection approaches based on payload inspection, fixedresolution spectral methods, or stationary traffic assumptions are often ineffective in encrypted and highly dynamic wireless environments. This paper proposes a Multiscale Adaptive Signal Processing (MASP) framework that combines Discrete Wavelet Transform (DWT)-based traffic decomposition with Zero Trust Architecture (ZTA) enforcement for efficient metadata-driven anomaly detection. Using Daubechies db4 wavelets and sliding-window packet-rate analysis, the framework extracts multiscale energy features that distinguish persistent attack traffic from legitimate transient bursts. The model incorporates multiscale feature extraction, anomaly scoring, and sigmoid-based risk mapping to support adaptive Zero Trust decision-making. Evaluation on the CAIDA DDoS Attack 2007 dataset with synthetic flooding augmentation demonstrates detection performance ranging from 88.7% to 100.0% TPR, 1.4% to 3.9% FPR, F1-scores up to 99.4%, and detection latency as low as 0.6 s. Comparative results show that MASP outperforms CUSUM, STFT-based, entropy-based, and Isolation Forest methods while maintaining computational efficiency suitable for real-time deployment in 5G edge-UPF environments and encrypted traffic monitoring.