Govt to use analytics, machine learning to detect objectionable video content

Aiming to check public disruptions by anti-social elements through objectionable videos via social media and chat apps, the government is working on creating a robust system based on machine learning and data analytics for detecting such content. The Ministry of Electronics and IT (MeitY) is working with research institutions on creating a system design for detecting objectionable video content using video analytics, a senior government official said. Through video analytics, various aspects such as when was the video uploaded or how was it circulated can be ascertained.

“The idea is to design a system that can analyse similarities in videos, determining the time of upload in case it can be found as a meta information in the host website, or designing a system that can determine whether an objectionable or obscene video is similar to the video in a given dataset,” the official explained.

Confirming the development, a top government official, who has worked on cybersecurity issues, said harassment on social media has become very common.

“Its not just objectionable videos showing sexual content, but many times anti-social elements upload videos related to violence against certain communities on social media sites like Facebook or video-sharing sites like YouTube, which become viral and lead to public disturbances. There are scores of such cases globally and in India where video content has been uploaded with a certain intent. The government wants to develop a mechanism through which it can track such content and if possible punish the perpetrators,”
he said.

This is besides the menace caused by videos circulating on the internet by various terror groups to create havoc among public or to brainwash people, especially the youth. The system that will be created will also track such videos, the official said.
The same official informed that the Centre for Development of Advanced Computing (CDAC) is working on the design and development of a video forensic tool for law enforcement agencies.

“This tool will have features like person tracking, person re-identification, skin tone analysis for categorisation, error analysis, video frame de-blurring and super resolution. This will help law enforcement and intelligence agencies analyse objectionable video content, the official said.

The government has the Information Technology Act 2000, which it amended in 2008 to include punishment for publishing or transmitting objectionable content. It also has legislations such as the Protection of Children from Sexual Offences Act and the Indecent Representation of Women (Prohibition) Act to restrict or prohibit publication of objectionable content in the electronic form.

“But many times the real perpetrators behind such acts go scot free due to lack of proper tools to identify them. In such case, laws sometimes become ineffective. Now data analytics, machine learning and artificial intelligence can help us track such anti-social elements,” the official explained.