csps (Civil Scam Prevention Software)

Inspiration

I’m from Nigeria, and it’s really hard to get a job anywhere in Europe or America because of the reputation Nigerians have. A majority of the country is unemployed and turns to scams as a way of living. It’s very hard for people like me who want to make a better living and are willing to put in the work for it. So I want to help prevent as many scams as I can.

What it does

CSPS is a forensic-grade, multi-vector impersonation detection engine built to catch sophisticated government and corporate scams that bypass traditional spam filters. It operates by analyzing sender domains and raw email headers for cryptographic signature failures (SPF, DKIM, DMARC), flagging risky payment methods like cryptocurrency or wire transfers, and leveraging NLP sentiment analysis to spot extreme pressure or urgency. The engine dynamically routes text through multi-vector context checks, such as civic permits, tax audits, and Microsoft 365 alerts, while automatically extracting Indicators of Compromise (IOCs) like malicious URLs and crypto wallets. Finally, it maintains strict digital forensics standards by generating SHA-256 cryptographic hashes of scanned texts and logging them securely into a local SQLite database.

How i built it

CSPS is built using a decoupled architecture consisting of a Python backend forensic engine and a client-side HTML and JavaScript dashboard. The backend is built as a FastAPI microservice that processes notice texts through specialized detection modules, including a domain analyzer, payment flagger, pressure scorer with sentiment analysis, plausibility checker, and an IOC extractor. For network security and evidence integrity, it includes header verification for cryptographic signatures and an SQLite database that logs every scan alongside SHA-256 cryptographic hashes of the input text. The frontend dashboard communicates with this Python API via REST to deliver real-time analysis.

Challenges we ran into

calibrating the risk scoring so it matched reality, not just theory. It was easy to build something that correctly caught an obvious scam, but the harder test was making sure a legitimate notice that’s a little pushy (real government offices do send reminders about deadlines and late fees) didn’t also get flagged as high risk. Getting false positives right mattered as much as catching the real scam pattern, since a tool that over-flags normal government mail loses people’s trust fast.

Accomplishments that we're proud of

Built a tool grounded in a real, currently active FBI-documented threat, it’s built directly from the specific fraud pattern investigators flagged this year, catching the exact combination of ways scammers are tricking people now.

What we learned

Building the detector taught me that a scam-detection tool is only as good as its false-positive rate, not just its catch rate. It’s not hard to build something that flags an obvious scam, the real difficulty is making sure it doesn’t also flag normal, legitimate communication.

What's next for csps (Civil Scam Prevention Software)

This is my Second Project as a Cyber Security major, I love what I’ve built and if it’s possible if the near future would like to see it grow into something respectable.

Built With

• python