Today, humans have become confronted with the hazards of robberies, assaults, and homicides, revealing that we are frequently the greatest menace to our own civilization. These conditions endanger those working alone at night in remote areas. The primary purpose of this project is to identify human screams for protection from unsafe surroundings or solitary confinement. The suggested methodology methodically develops a deep neural network model for classifying instances as screams or non-screams. The model captures variables including chroma, mel spectrogram, and mel-frequency cepstral coefficients (MFCC). The suggested strategy achieves 96.6% accuracy in categorizing auditory events relevant to potential threats, demonstrating its usefulness. Moreover, the system protects individuals by automatically alerting registered contacts through various communication channels.