In this paper, we have proposed a novel architecture for citizen security using deep learning based edge computing device. The proposed system comprises of an end-device which can be connected to every street light pole in an area. The gateway device can be connected at a large distance from the end-devices. LoRa physical layer is used as the main communication interface which enables large distance transmission. CNN based algorithm is used to identify the expression of the person who pressed the panic button on the device to reduce false alarm. A very low complex Convolutional Neural Network (CNN) architecture has been developed, to recognize the expression on the end device itself. A data logging server is used for logging the node-id everytime it is pressed. An android application is also developed which shows the real-Time location of the pressed node and provides navigation routes to reach that node. © 2019 IEEE.