.png&w=256&q=75)
Shreya Prasad@Shreya_07
1
Events attended
1
Submissions made
Looking for experience!
Socials
🤝 Top Collaborators
.png&w=640&q=75)
.png&w=640&q=75)
🤓 Latest Submissions
SecureC
SecureC is an AI-native Web Application Firewall designed for the agentic era—where the attack surface is a conversation, not a URL. The Problem: Traditional WAFs rely on static regex rules that fail against dynamic prompt injection attacks, jailbreaks, and AI-specific threats like PII leakage in model outputs. As enterprises deploy AI agents, they need security tooling that understands natural language context. Our Solution: SecureC implements a triple-layer defense: Input Guard — Detects prompt injection, role override, and delimiter abuse before inputs reach AI agents Output Guard — Scans AI responses for PII, API keys, and sensitive data with automatic redaction Behavior Guard — Monitors agent behavior against scoped permissions, enforcing least-privilege at the AI layer Technical Innovation: Our local ML analytics engine computes Shannon entropy and security keyword frequency analysis with zero external API dependencies. This hybrid approach combines the reasoning power of LLMs with deterministic statistical signals for robust threat detection. Multi-Agent Pipeline: Five specialized agents (Threat Modeler, Security Auditor, SOC Intelligence, Remediation Engineer, Risk Strategist) analyze artifacts and synthesize a GO/NO-GO deployment decision. Each agent has strict scope boundaries—the Remediation Agent suggests fixes but cannot auto-apply them. Enterprise-Ready: Supabase integration for persistent vulnerability logging, Slack webhooks for real-time SOC alerts, and an escalation policy that surfaces agent disagreement for human review. Built with FastAPI, React, and OpenRouter. Defensive programming throughout—graceful fallbacks for malformed LLM responses, confidence clamping, Pydantic schema validation.
7 Feb 2026
