PrivacyEye: Controlling Harmful Multimedia Sharing Among Children

The increasing use of electronic devices and online applications among children in the UK has raised significant concerns about their online safety. Nearly 90% of children aged 0-18 go online daily, with those aged 5-15 spending an average of 5 hours and 24 minutes on social media daily, despite school commitments. Popular platforms like Facebook, Instagram, and Snapchat are widely used, with 79% of children aged 3-17 engaging in messaging or video calls. Alarmingly, 67% of UK parents express concerns about their children's online activities, as cybercrimes targeting children, such as exposure to harmful content, cyberbullying, and online exploitation, continue to rise. Reports indicate that one online abuse offence against a child was recorded every 16 minutes in England and Wales between April 2019 and January 2020\. Key factors contributing to these risks include children accepting friend requests from strangers, with 682,000 children engaging in online conversations with unknown individuals over a year. Additionally, 11% of children aged 13-15 reported receiving sexual messages, increasing their vulnerability to bullying, isolation, and sexual abuse. Perpetrators often pressure victims into producing explicit content, threatening to share it with their networks. These trends highlight the urgent need for effective solutions to protect children online, ensuring their safety and compliance with regulations like GDPR. PrivacyEye, an AI-driven cybersecurity toolkit, addresses these challenges by detecting and masking of harmful content, delivering a scalable and cost-efficient approach to safeguarding children in digital environments. This project introduces a privacy-focused toolkit designed to prevent unintentional capture of individuals, body parts, or objects during video calls. By integrating with video conferencing apps or operating systems, it offers real-time control over camera feeds, enhancing privacy and security. The toolkit combats risks like child exploitation and cyberstalking, particularly in environments where children use devices unsupervised. It also supports BYOD policies, ensuring safer video communication and preventing unauthorized sharing of sensitive content. PrivacyEye's core features include: (1) advanced machine learning algorithms to identify and mask harmful content during video calls, ensuring privacy and security; and (2) customizable privacy zones that automatically blur or obscure specific areas or objects, preventing unintentional exposure. PrivacyEye enhances accuracy and transparency with explainable AI (XAI) and uses model quantisation and pruning to improve scalability by reducing the computational load and speeding up multimedia analysis on Android devices. A parental control dashboard provides real-time alerts, usage analytics, and content review options, empowering parents to manage their children's online activities.