Children suffering from Autism Spectrum Disorder (ASD) frequently experience sudden panic attack episodes in their parents' or caregivers' absence, leading to feelings of insecurity and isolation. To counteract this problem, we have designed an Internet of Things (IoT)-based smart monitoring system that can detect panic attack episodes in real time, automatically alarming parents, caregivers, or teachers promptly. The system consists of ESP32 and ESP32-CAM in combination with sensors for temperature, humidity, light, and sound measurement, while an Arduino Uno controls actuators such as a fan, LED, and speaker to allow prompt responses. For our purpose of panic attack detection, we used MediaPipe Pose for pose and landmark detection in combination with a TensorFlow Lite neural network architecture, achieving robust classification of abnormal behaviors by more than 92 percent in our tests. When a panic attack is sensed, the system automatically triggers environmental changes and sends an alert via email. Simultaneously, sensor data and live video can be monitored in real-time through the web dashboard. This inexpensive, real-time, and scalable solution improves child safety, informs caregivers, and allows early intervention in such places as schools, therapy institutions, and home settings.