AwasScam is a specialized cybersecurity tool designed to protect Malaysians from the rising tide of digital fraud. Unlike generic AI, AwasScam is fine-tuned with deep knowledge of the Malaysian landscape, recognizing local bank patterns (Maybank, CIMB), government agencies (LHDN, KWSP, PDRM), and local dialects (Malay, Manglish). The app features a multimodal "Threat Scanner" that analyzes both text messages and screenshots. Beyond just detection, it provides a "Cybersecurity Threat Report" that simulates the scammer's next move and explains the scam's progression step-by-step to educate users on psychological manipulation tactics.
Architecture & Tech Stack:
- Frontend: Built as a modern Single Page Application (SPA) using React 19 and TypeScript for robust type-safety.
- AI Integration: Leverages the Google Gemini SDK (@google/genai) for real-time, multimodal inference (Text + Image).
- Styling: Tailwind CSS with a custom "Cybersecurity Dashboard" theme, utilizing a dark-navy palette and high-contrast indicators.
- Animations: Motion (Framer Motion) for smooth UI transitions, progress bar animations, and a custom "laser-scan" effect during analysis.
- Deployment: Hosted on Google Cloud Run, ensuring high availability and fast response times for users across Malaysia.
The prompt engineering process focused on creating a high-precision System Instruction that locks Gemini into a "Malaysian Cybersecurity Expert" persona.
- Contextual Guardrails: The prompt includes a knowledge base of specific Malaysian scam tactics (e.g., fake PosLaju delivery, NSRC 997 impersonation, and LHDN tax refunds).
- Multimodal Logic: Optimized to handle raw text and OCR-extracted text from images simultaneously, ensuring accurate analysis of WhatsApp/SMS screenshots.
- Structured Output (JSON): Enforces a strict JSON schema to ensure the frontend can reliably parse data for UI components like risk badges, probability bars, and the "Scam Progression" timeline.
- Educational Simulation: Designed to generate a "Next Move" simulation, which helps users visualize the scammer's manipulation tactics before they happen.
Log in or sign up for Devpost to join the conversation.