AI-Driven Cyber Risk Assessment
Budget: ₹250,000 – ₹500,000 INR
I’m expanding InsureNet with an AI-powered module focused exclusively on assessing cyber risks for our business customers. The goal is to combine three proven methods—vulnerability scanning, threat-intelligence analysis, and security audits—into a single, automated workflow that produces a clear risk score and supporting evidence that underwriters can trust.
What I need from you
• Build or integrate reliable vulnerability scanners (OWASP ZAP, Nessus, or a comparable open-source tool) that can be triggered via API and return normalized findings.
• Pull real-time threat-intelligence feeds (STIX/TAXII, CVE databases, commercial feeds if you have access) and correlate them with the scanner output.
• Design an ingest mechanism for security-audit data so that audit controls, policy gaps, and configuration drift influence the overall score.
• Develop the machine-learning logic—ideally in Python with scikit-learn, TensorFlow, or PyTorch—that weights each data source, explains the score, and improves itself over time.
• Expose everything as a documented REST or GraphQL service deployable through Docker (Kubernetes-ready).
• Include unit tests, a small synthetic dataset for demo purposes, and concise setup instructions.
Acceptance criteria
1. Given sample scan output, threat-intel events, and audit findings, the service returns a numeric risk score plus a JSON explanation of contributing factors.
2. Average processing time per assessment is under 60 seconds on a standard cloud VM.
3. All components run in a single docker-compose stack and pass the provided test suite.
If you’ve built similar scoring engines, or have hands-on experience with cyber-insurance data, I’d love to see your approach and timeline.
What I need from you
• Build or integrate reliable vulnerability scanners (OWASP ZAP, Nessus, or a comparable open-source tool) that can be triggered via API and return normalized findings.
• Pull real-time threat-intelligence feeds (STIX/TAXII, CVE databases, commercial feeds if you have access) and correlate them with the scanner output.
• Design an ingest mechanism for security-audit data so that audit controls, policy gaps, and configuration drift influence the overall score.
• Develop the machine-learning logic—ideally in Python with scikit-learn, TensorFlow, or PyTorch—that weights each data source, explains the score, and improves itself over time.
• Expose everything as a documented REST or GraphQL service deployable through Docker (Kubernetes-ready).
• Include unit tests, a small synthetic dataset for demo purposes, and concise setup instructions.
Acceptance criteria
1. Given sample scan output, threat-intel events, and audit findings, the service returns a numeric risk score plus a JSON explanation of contributing factors.
2. Average processing time per assessment is under 60 seconds on a standard cloud VM.
3. All components run in a single docker-compose stack and pass the provided test suite.
If you’ve built similar scoring engines, or have hands-on experience with cyber-insurance data, I’d love to see your approach and timeline.