Biometrics and PINs are frequently used for authentication in secure access systems, but they both have issues with spoofing resistance, revocability, and deployment costs. Dynamic signatures are inherently flexible and revocable, in contrast to fingerprints or facial biometrics, which cannot be changed once compromised. An inertial measurement unit (IMU) built into an ESP32-based electronic pen is used to record in-air dynamic signatures, which are the basis of the privacy-preserving authentication framework presented in this paper. A lightweight 1D-CNN model trained in Edge Impulse is used by the ESP32 to locally process 6-axis motion data, allowing for real-time user identification independent of the cloud. The system ensures low-latency and secure peer-to-peer communication by sending an AES-encrypted authentication token over ESP-NOW after verification. The suggested approach lowers the possibility of replication or observation attacks because it leaves no visible trace, in contrast to traditional signatures. Strong resistance to spoofing, low latency, and consistent accuracy are demonstrated by experimental evaluation. The method provides an affordable edge-compute solution that can be used in identity-based access control systems, smart infrastructure, and secure Internet of Things environments.