Skip to content

Tharun8688/PiShieldd

Repository files navigation

# Pi Shield - The Information Sentinel Pi Shield is a multi-platform, AI-powered tool designed not just to fact-check, but to educate Indian users on how to identify manipulative content, fostering long-term digital literacy. ## Core Concept Pi Shield operates on a simple principle: "Teach a user to spot a lie, and you protect them for a lifetime." Instead of just giving a "true" or "false" verdict, it analyzes content (text and images) and provides a Credibility Score along with a simple, educational breakdown of the persuasion and manipulation techniques being used. It will be primarily accessible via a WhatsApp chatbot to meet users on the platform where misinformation is most rampant. ## Key Features 1. **Multi-Modal Analysis** Users can forward WhatsApp messages (text), images, or paste links into a web app. Pi Shield analyzes all of these formats. 2. **The Deconstruction Engine (Core AI Feature)** This is the heart of the solution. Using Generative AI, the engine doesn't just check facts; it identifies and explains common misinformation tactics: * **Emotional Manipulation:** Detects emotionally charged language (fear, anger, outrage, urgency) designed to bypass critical thinking. * **Source Analysis:** Checks if credible sources are cited. It flags claims that are unsourced or linked to known propaganda sites. * **Image Forensics:** Performs a reverse image search to check if an image is old, used out of context, or digitally altered. It also flags tell-tale signs of AI-generated images. * **Scam Detection:** Recognizes patterns common in financial scams, such as promises of guaranteed high returns, urgent calls to action, and requests for personal information or small payments. 3. **Credibility Score & Educational Cards** After analysis, the tool provides a simple, color-coded credibility score (e.g., High Risk, Be Cautious, Likely Credible). More importantly, it generates easy-to-read "Educational Cards" that explain its reasoning in plain language (supporting multiple Indian languages). *Example Card:* "Be Cautious 🤔 This message scores low on credibility. Here's why: * **Urgent Language:** It uses words like 'act now' to create panic. * **No Source:** It makes a big claim but doesn't say where the information came from. * **Old Image:** The photo is from a different event five years ago." 4. **"Learn Mode"** An interactive section with short, gamified tutorials on topics like "How to spot a deepfake," "Understanding propaganda," and "Recognizing phishing scams." ## User Experience: The WhatsApp Flow * **Submission:** A user receives a suspicious message or image on WhatsApp. They forward it directly to the "Pi Shield" contact number. * **Acknowledgement:** The bot immediately replies: "Thanks! I'm analyzing this for you. This should only take a moment... 🧐" * **Response:** Within seconds, the bot sends back the concise Educational Card with the credibility score and the key red flags. It includes a button: "Learn more about these tricks." * **Deep Dive (Optional):** If the user clicks the button, the bot sends a more detailed explanation of the manipulative techniques identified, helping the user understand the why behind the rating. ## Technical Architecture on Google Cloud This solution is designed to be scalable, secure, and cost-effective using a serverless architecture on Google Cloud. * **User Interface (Entry Point):** * **WhatsApp Bot:** A third-party gateway like Twilio integrates with a backend service. * **Web App:** A simple frontend hosted on Firebase Hosting. * **Backend & Orchestration:** * **Cloud Functions:** The primary logic is handled by these event-driven functions. A function is triggered when a message is received from the WhatsApp gateway. * **Cloud Run:** Hosts the main application container that orchestrates the different AI calls. * **Generative AI Core (Vertex AI):** * **Gemini Pro Model:** This is the core engine. Cloud Functions will call the Gemini API with a specifically engineered prompt to perform the "Deconstruction." * *Example Prompt to Gemini:* "Analyze the following text for misinformation techniques. Identify emotional language, unverifiable claims, and indicators of a scam. Return your findings as a structured JSON object with a confidence score for each technique detected." * **Vertex AI Vision:** Used for Optical Character Recognition (OCR) to extract text from images and to analyze images for signs of manipulation or AI generation. * **Data & Knowledge Base:** * **Firestore:** A NoSQL database to store user interaction data, known scam formats, and a curated list of credible Indian news sources for cross-referencing. * **Google Custom Search API:** Programmatically performs reverse image searches and checks claims against trusted news websites. ## Why This Solution is Innovative * **Focus on Education, Not Just Verdicts:** It directly tackles the core objective of creating a more critical and informed citizenry by teaching the methods of misinformation. * **Meets Users Where They Are:** The WhatsApp-first approach is tailored to the Indian context, making the tool incredibly accessible without needing to download another app. * **Contextual, "Just-in-Time" Learning:** The educational component is delivered at the exact moment of curiosity and doubt, making the lesson far more impactful. * **Proactive Defense:** By learning to recognize the patterns, users become proactively resistant to future misinformation, reducing the overall spread of harmful content over the long term. # PiShieldd

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published