Abstract Our project aims at detecting any abnormal behaviour in CCTV footage to solve many accidents, crimes, security issues, etc. CCTV footage gives the best information to detect any real-time abnormal behaviour and hence the CCTV footage is incorporated in the proposed project. Dataset is prepared by adding abnormalities by the use of Gaussian blur/noise, to replace the training set of normal images with abnormal images and to make sure that there is no bias in the dataset. To get better performance from our dataset, feature normalization is harnessed. Moreover, we used a Convolutional Neural Network (CNN) as our model to convolute and predict the images in our dataset, which achieves better performance than any other neural network existing today. The proposed neural network performs optimization based on Nadam and the experimental outcomes show that the proposed neural network offers superior results compared to that of its peers. Therefore, it has been observed that the proposed neural network obtains better accuracy than conventional neural networks.