DETECTION OF ILLICIT MESSAGES IN TWITTER USING SUPPORT VECTOR MACHINE AND VGG16

Abstract

Human trafficking is a serious issue all around the world. Millions of men, women, and children are drawn into illicit activities and become victims as a result. People smuggling is prevalent in many nations, particularly among children and teens under the age of 14. Social media platforms serve as hurdles to the proliferation of these crimes through the internet environment, where subtle messages urge the use of unlawful services. It is critical to automatically detect these communications in real time. These aid government officials in preventing social networking crimes.

Machine Learning and Deep Learning Algorithms are used to detect, preprocess, and classify suspicious tweets based on gender and age group, as well as photos from the tweets.

Convolution Neural Networks (CNN) is used in the current techniques for detecting unauthorized tweets and photos. Although the CNN algorithms is effective at identifying human photographs, they are unable to locate all real-time tweets. When compared to VGG16, the picture information used is not considered adequately. To identify unlawful messages, CNN is utilized, although with lower accuracy rates.

The suggested approach extracts images of people and hashtags or language relevant to minors (under 14 years) in real time and preprocesses it to remove misspelt and noisy data, after which the tweets are classed as suspicious or not. Support Vector Machine (SVM) and VGG16 are used to classify the gender and age of the photos in the tweets. Machine Learning and Deep Learning Algorithms such as SVM and VGG16 are used to identify, preprocess, classify, and determine the accuracy of these text and image tweets in order to remove misspelt and noisy data.