TRU | Devpost

home page
trust score
live_feed
workflow

Inspiration

The idea for TRU (Trust & Reputation Unit) came from a simple but worrying reality:
in the age of AI-generated content, lies spread faster than truth.

Fake news, phishing scams, and manipulated content can go viral within minutes, while verification often takes hours or days. During critical moments—elections, pandemics, or financial crises—this delay causes real-world harm. Existing fact-checking systems are mostly centralized, slow, and vulnerable to censorship or external pressure.

We asked one core question:

“How can trust on the internet be verified instantly, transparently, and without relying on a single authority?”

That question became the foundation of TRU.

What We Built

TRU is a multi-layer trust verification system that combines:

AI-powered content analysis for real-time misinformation detection
Blockchain-based immutability for tamper-proof verification records
Community-driven reporting with economic incentives to prevent abuse

At its core, TRU works as a digital notary for online content—verifying claims, assigning trust scores, and anchoring proof on-chain.

How We Built It

AI Verification Engine

We developed an AI pipeline using a fine-tuned DistilRoBERTa-based transformer trained on verified fake-news datasets.
Instead of a simple true/false output, the system provides:

A Trust Score (0–100)
A confidence value
Explainable signals highlighting why a claim may be unreliable

This ensures transparency and avoids blind reliance on AI decisions.

Backend & APIs

The backend is built with FastAPI, enabling fast and asynchronous request handling.
Core endpoints include:

/verify – analyze content and generate trust scores with cryptographic hashes
/feed – aggregate live cybersecurity threats via RSS feeds
/report – allow community reporting of scams and misinformation

In-memory caching is used to maintain low-latency responses.

Blockchain Layer

TRU stores only essential metadata on-chain:

SHA-256 content hash
Verification verdict
Timestamp

This approach preserves privacy while ensuring censorship-resistant and auditable verification.

Trust Economics

To prevent spam and coordinated attacks, TRU uses stake-based reporting:

Users stake tokens when submitting reports
Accurate reports are rewarded
False or malicious reports lose their stake

This aligns incentives and discourages manipulation.

Challenges & Learnings

Balancing speed and accuracy required confidence intervals and an uncertainty zone
Preventing AI overconfidence meant never claiming absolute certainty
Designing against spam and brigading required economic deterrents
Blockchain was used carefully as a verification anchor, not a storage layer

We learned that trust is not binary—it is probabilistic, contextual, and adversarial.

Impact & Vision

TRU is built to scale beyond a hackathon project.
Our vision is to make trust verification a public digital utility for users, platforms, and institutions.

In a world where content is created instantly, truth must be verified even faster.

That is the mission of TRU.