TLDR: SafeSpace is a web application that integrates three key modules to enhance digital safety and emotional well-being: toxicity detection in chats and screenshots using NLP, a configurable safety ping system for emergency alerts with live location, and a reflective questionnaire to assess relationship health. Developed by Kayenat Fatmi and Mohammad Abbas, the platform aims to provide a holistic solution to online harms by combining detection, protection, and reflection, addressing the fragmentation of existing safety tools. It demonstrates high precision in toxicity detection and reliable safety alerts, with future plans for mobile deployment and advanced alert mechanisms.
In our increasingly digital world, connecting and sharing online has become second nature. However, this digital shift also brings significant risks, including exposure to online harms like toxicity, manipulation, and grooming. These issues can have serious emotional and safety consequences, particularly for vulnerable individuals. While many tools exist to tackle these problems, they often operate in isolation – some detect abusive content, others issue safety alerts, but rarely do they combine digital safety with emotional well-being.
Addressing this critical gap, researchers Kayenat Fatmi and Mohammad Abbas from Jamia Hamdard University have introduced SafeSpace, an innovative web application designed to offer a unified solution for digital safety and emotional health. SafeSpace integrates three core modules into a single, user-friendly platform, moving beyond the fragmented approach of existing tools.
The Three Pillars of SafeSpace
SafeSpace is built around three distinct yet interconnected functionalities:
1. Chat Toxicity Analysis: This module uses advanced Natural Language Processing (NLP) models, including Google’s Perspective API, to detect abusive, manipulative, or emotionally harmful language in chats and even screenshots. Users can input text or upload images, and the system provides a report highlighting toxic phrases and overall toxicity levels. This helps users identify and understand potentially harmful interactions in their private digital conversations, an area often overlooked by public platform moderation tools.
2. Configurable Safety Ping System: Designed for proactive protection, this system allows users to set custom check-in intervals. If a user misses a scheduled check-in, SafeSpace automatically triggers an emergency alert. These alerts are sent to predefined emergency contacts via SMTP-based emails, including the user’s live location (longitude and latitude). The system also supports manual SOS alerts, ensuring help can be summoned quickly when needed. Firebase is used for efficient alert management, and offline caching ensures alerts are delivered even if connectivity is temporarily lost.
3. Reflective Questionnaire: Recognizing the importance of emotional well-being, SafeSpace includes a reflective questionnaire. This tool evaluates relationship health and emotional resilience based on psychological assessment scales. Users answer curated questions on communication, trust, and emotional well-being, and the system provides personalized feedback categorized as “Healthy,” “Needs Reflection,” or “Unhealthy,” along with supportive advice. This module empowers users to gain insights into their emotional state and relationship dynamics, fostering long-term resilience.
A Holistic Approach to Digital Well-being
Unlike many existing safety applications that primarily focus on physical protection or panic buttons, SafeSpace uniquely incorporates digital conversations and emotional resilience as integral parts of user safety. The platform emphasizes privacy, ensuring that sensitive chat inputs are processed transiently and not stored permanently. Its modular architecture is designed for usability, privacy, and scalability, making it accessible for non-technical users while allowing for future enhancements.
The implementation of SafeSpace involved a Flask-based backend, Firebase for alert management, and machine learning APIs for NLP. The system’s architecture includes a responsive user interface, an application logic layer to manage requests and workflows, backend and storage services for user settings and alert history, and robust notification services.
Also Read:
- AgentScope 1.0: Empowering Developers to Build Advanced AI Agent Applications
- Unpacking AI’s Grasp of Human Reasoning Styles in Social Games
Promising Results and Future Vision
Experimental evaluations of SafeSpace have shown impressive results. The toxicity detection module achieved approximately 93% precision, with a false positive rate of less than 5%. The safety ping system demonstrated 100% reliability in emulator tests, with alerts triggering within 30 seconds of a missed check-in. The questionnaire scoring showed a 92% alignment with manual scoring, and the system received a high usability rating of 4.6/5 from peer surveys.
While SafeSpace excels in accuracy, reliability, and user-centric design, the researchers acknowledge certain limitations, such as challenges with sarcasm and cultural nuances in toxicity detection, and current restrictions on direct integration with third-party messaging platforms. The reliance on email for alerts also highlights a need for SMS or call fallback in real emergencies.
Looking ahead, the future development of SafeSpace envisions deployment as a dedicated mobile application for broader accessibility, advanced alert mechanisms including SMS and automated phone calls, multilingual support, and sentiment trend visualization dashboards. The researchers also plan for human-in-the-loop validation with mental health professionals and collaborations with law enforcement and advocacy groups to extend its societal impact.
SafeSpace represents a significant step forward in digital safety, offering a proactive, user-centric platform that merges detection, reflection, and protection to foster holistic digital resilience. You can read the full research paper here: SafeSpace Research Paper.
