alovia ai

Alovia Shield — Block AI crawlers before they steal your content

Comment

Inspiration

Write a project "About" section for Alovia Shield formatted in Markdown with LaTeX math support. Save to content/about-alovia.md

Use this structure and these real details:

Inspiration

  • Attended an event where I learned AI companies were scraping web content without permission or compensation to creators
  • As a CS student and content creator I felt the problem personally
  • Realized existing tools like robots.txt and Cloudflare weren't built to stop AI crawlers specifically
  • Decided to build the solution instead of waiting for someone else to

What I Learned

  • How AI crawlers work differently from regular bots — they ignore robots.txt, spoof user agents, mimic real browsers
  • Deep dive into TLS/JA3 fingerprinting, behavioral analysis, header anomaly detection
  • Built a 6-check async concurrent detection engine from scratch
  • How to measure false positive rates and tune detection thresholds
  • The math behind detection accuracy — include this LaTeX formula: Accuracy = (TP + TN) / (TP + TN + FP + FN) And pass rate formula: PassRate = TestedPassed / TotalTests * 100 = 37/40 * 100 = 92.5\%

How I Built It

  • Stack: Next.js, TypeScript, Supabase, custom detection middleware
  • 6 concurrent detection checks: bot signature matching, user-agent analysis, request pattern detection, header anomaly scanning, behavioral fingerprinting, TLS/JA3 fingerprint analysis
  • Built a full test suite: 40 tests across LOW/MEDIUM/HARD/EXPERT difficulty levels
  • Watermarking layer: zero-width characters, homoglyph substitution, LSB steganography, DCT watermarking, perceptual hashing
  • Iterated using Claude Code for automated testing and report generation
  • Custom headers: Server: Alovia, X-Protected-By: Alovia Shield

Challenges

  • False positives: legitimate traffic getting blocked — resolved by adding 137.184.39.84 to SHIELD_HINTS whitelist
  • Getting EXPERT level bots (full browser mimicking) to trigger correctly
  • Balancing detection sensitivity vs blocking real users
  • Building solo as a CS student while managing coursework and HackHayward
  • DMARC configuration still in progress (p=quarantine on IONOS DNS)

Results

  • 40 tests, 92.5% pass rate
  • 0 false positives after tuning
  • 6-check async engine running concurrently
  • Built and tested solo in pre-launch phase

Make it sound authentic, first person, builder voice. Not a press release — a real founder story. Include the LaTeX formulas properly formatted with $$ delimiters. Keep total length around 400-500 words.

What it does

How I built it

Challenges I ran into

Accomplishments that I'm proud of

What I learned

What's next for alovia ai

Built With

Share this project:

Updates