Child Safe Browser Extension: A Browser Extension to Detect Adultery and Violent Content to Make Safer Web for Children

Abstract

The world is changing with the pace of information technology revolution and now a-days anybody can access to the internet including the children. Since birth These 21st century children are able to access to the di erent types of websites on the internet because of their accessible devices but not every time the internet websites are child friendly and children can view some violent and adultery images which can a ect their childhood along with their mind. Furthermore, most of the people are having business from websites what is called E-commerce nowadays but the issue is that the business persons are adding ads to their websites for the pro ts. However, sometimes these ads can be violent and also can contain adultery images which can be viewed by a child. To stop, viewing these types of images the main way can be stopping the access of the images from a child's device which can be done by adding extension to their devices. Then parents can feel relief and children can gain their knowledge by using the great side of the internet. So, to not access the images a lter should be made but the main challenge is the making of lter which will know a speci c work to maintain the extension. Also, none of the model can give 100 percent. However, when an algorithm gets heavier then it will work slowly and it will have an impact on the browser speed so to avoid this issue the model need to make a lighter algorithm. Here, to make this extension happen and e client it will need image processing and machine learning so that it can help to make a virtual surveillance for the children and can make their mind fresh and creative. Our system will be hosted in a cloud space and will be called by the extension with the help of an API. In our system we used beautifulsoup4 for image scrapping from web page. Augmentor is for the data augmentation. For a faster machine learning PIL is our choice. And Tensorow performed extremely well in our Convolution Neural Network.