1
1
Philippines
1 year of experience

Overview In the year 2026, deepfakes have become indistinguishable from reality, threatening democracy, global finance, and the judicial system. DeepShield, developed by The Moon Project, is a sophisticated defense layer that utilizes a Multi Agent Architecture to identify synthetic media that 95 percent of humans would believe to be real. It serves as an essential tool for banks, news agencies, and governments to prevent hundred million dollar frauds and large scale misinformation. The 4 Agent Multi Agent System DeepShield employs an Orchestrator Agent that synthesizes data from four specialized AI agents to reach a final high confidence verdict: Agent 1 Facial Authenticity Analyzer: Performs deep level scans of eye reflections, skin texture, and micro expressions to find subtle noise artifacts and shadow vectors left by AI generators. Agent 2 Liveness Detector: Verifies if the subject is a living person through Remote Photoplethysmography to detect blood flow in the face and natural micro movements that synthetic media cannot replicate. Agent 3 Behavioral Anomaly Detector: Analyzes lip sync accuracy, blink rate patterns, and gesture timing to flag unnatural movements typical of AI. Agent 4 Source Verification: Performs a forensic audit for cryptographic watermarks and camera hardware signatures to ensure the video was captured by a physical device rather than generated in software. Real World Impact and Security DeepShield is built for critical scenarios such as preventing Election Fraud, securing Bank Transactions against CEO impersonations, and validating legal evidence in court. To ensure privacy, it integrates with Lobster Trap protocols which enforce strict data policies where agents analyze biometric data but are prohibited from storing or sharing it with third parties. Using a tech stack including Gemini Pro and Gemini Vision, DeepShield is the ultimate shield for a world where AI must verify the truth.
19 May 2026