BAAR Logo

BAAR3-Captcha-Guard (AI-Resistant Multimodal CAPTCHA Engine)

🌐 Official Website: https://baar.uk/model/
πŸš€ Model Hub: aixk/BAAR3-Captcha-Guard
πŸ“„ License: Non-Commercial Only (Contact for Commercial Pricing)


Captcha Guard Overview


πŸ“Œ Model Overview (λͺ¨λΈ κ°œμš”)

BAAR3-Captcha-Guard is an ultra-fast, edge-native multimodal CAPTCHA defense engine designed to counter Vision LLMs (GPT-4o, Claude Computer-Use), automated headless scripts (Puppeteer, Playwright), and replay injection attacks. By fusing visual challenge canvases with user interaction dynamics, the model delivers zero-trust bot discrimination with sub-millisecond edge and WASM latency.

BAAR3-Captcha-GuardλŠ” μ΅œμ‹  Vision AI μ—μ΄μ „νŠΈ(GPT-4o, Claude Computer-Use λ“±), λΈŒλΌμš°μ € μžλ™ν™” 슀크립트(Puppeteer/Selenium), λ¦¬ν”Œλ ˆμ΄ λ³€μ‘° 곡격을 μ°¨λ‹¨ν•˜κΈ° μœ„ν•œ μ΄ˆκ²½λŸ‰ μ—£μ§€ 인라인 λ³΄μ•ˆ μ—”μ§„μž…λ‹ˆλ‹€. μΊ”λ²„μŠ€ 이미지 νŒ¨ν„΄κ³Ό μ‚¬μš©μž μ‘°μž‘ 동역학을 κ²°ν•© νŒλ³„ν•˜μ—¬ 정상 μœ μ € κ²½ν—˜μ„ ν•΄μΉ˜μ§€ μ•ŠμœΌλ©΄μ„œ 0.4ms λ‚΄μ™Έμ˜ κ·Ήμ €μ§€μ—° ν™˜κ²½μ—μ„œ μ§€λŠ₯ν˜• 봇을 μ‹€μ‹œκ°„ μ‹λ³„Β·μ°¨λ‹¨ν•©λ‹ˆλ‹€.


πŸ›‘οΈ Defense & Performance Benchmarks (μ‹€μ „ λ°©μ–΄ μ§€ν‘œ)

Validation Setup: 10,000 Verified Test Vectors | Runtime: Edge INT8 Quantized Engine

Class / Category Classification Type Detection Rate (μ°¨λ‹¨μœ¨ / ν†΅κ³Όμœ¨) Error Metric (였λ₯˜μœ¨) Primary Threat Scenarios
Class 0: 정상 인간 (Human) Benign User 99.82% ν†΅κ³Όμœ¨ FPR: 0.18% μžμ—°μŠ€λŸ¬μš΄ 인체 μ‘°μž‘ ꢀ적, 정상 ν„°μΉ˜/마우슀 μΈν„°λž™μ…˜
Class 1: 맀크둜/슀크립트 (Macro) Threat Flow 100.00% μ°¨λ‹¨μœ¨ FNR: 0.00% Puppeteer/Playwright μžλ™ν™”, 등속도 및 λ² μ§€μ–΄ 슀크립트
Class 2: Vision AI (LLM Agent) Threat Flow 99.88% μ°¨λ‹¨μœ¨ FNR: 0.12% λ©€ν‹°λͺ¨λ‹¬ LLM ν™”λ©΄ μΆ”λ‘  μ—μ΄μ „νŠΈ, μŠ€ν… 보정 μ‘°μž‘κΈ°
Class 3: λ¦¬ν”Œλ ˆμ΄ 곡격 (Replay) Threat Flow 99.95% μ°¨λ‹¨μœ¨ FNR: 0.05% 이전 μ„Έμ…˜ ꢀ적 볡제 μ£Όμž…, νŒ¨ν‚· 타이밍 μ‘°μž‘ μŠ€ν‘Έν•‘

⚑ μ’…ν•© λ°©μ–΄ μ§€ν‘œ μš”μ•½

  • μ‹€μ „ μ’…ν•© 봇 μ°¨λ‹¨μœ¨ (Attack Block Rate): 99.94% (미탐λ₯  FNR: 0.06%)
  • 정상 μ‚¬μš©μž ν†΅κ³Όμœ¨ (Benign Pass Rate): 99.82% (μ˜€νƒλ₯  FPR: 0.18%)
  • Vision-LLM μ—μ΄μ „νŠΈ μ°¨λ‹¨μœ¨: 99.88%
  • μ΅œμ’… 검증 볡합 점수 (Overall Score): 99.88 / 100.00

⚑ Inference Speed & Runtime Benchmark (μΆ”λ‘  벀치마크)

Runtime Engine Single Inference Latency Model Binary Size Runtime Characteristics
CDN Dynamic INT8 ONNX 0.42 ms ~320 KiB Edge Worker(Cloudflare/Fastly), WASM λΈŒλΌμš°μ € 인라인 배포 μ΅œμ ν™”
Standard ONNX (CPU) 0.85 ms ~1.25 MiB Node.js, Python, Go API κ²Œμ΄νŠΈμ›¨μ΄ λ°±μ—”λ“œ ν˜Έν™˜
PyTorch SafeTensors 1.10 ms ~1.28 MiB 고속 λ‘œλ”© 및 κ°€μ€‘μΉ˜ λ³΄μ•ˆ 검증 λ„€μ΄ν‹°λΈŒ ν…μ„œ

μ΄ˆμ €μ§€μ—° CDN μ—£μ§€ ꡬ동: 320KB μˆ˜μ€€μ˜ μ΄ˆκ²½λŸ‰ INT8 ONNX λ°”μ΄λ„ˆλ¦¬λ₯Ό 톡해 λΈŒλΌμš°μ € μ„Έμ…˜ 검증 λ˜λŠ” ν”„λ‘ νŠΈμ—”λ“œ CDN μ—£μ§€ λ‹¨μ—μ„œ 0.5ms μ΄λ‚΄λ‘œ 즉각적인 νŒλ³„μ„ μˆ˜ν–‰ν•©λ‹ˆλ‹€.


πŸ“¦ Export Formats & Deployment Artifacts (배포 파일 ꡬ성)

  • model.safetensors: 고속 λ‘œλ”© 및 λ³΄μ•ˆ 검증을 ν†΅κ³Όν•œ 순수 κ°€μ€‘μΉ˜ ν…μ„œ 파일
  • model.onnx: 크둜슀 ν”Œλž«νΌ ν‘œμ€€ ONNX λͺ¨λΈ κ·Έλž˜ν”„
  • model_cdn_int8.onnx: CDN Edge Worker 및 WASM ν™˜κ²½μš© INT8 μ΄ˆκ²½λŸ‰ λ°”μ΄λ„ˆλ¦¬ (~320 KiB)
  • config.json: μž…μΆœλ ₯ 규격 및 클래슀 λ ˆμ΄λΈ” λ§€ν•‘ 메타데이터

πŸ“„ License & Commercial Terms (λΌμ΄μ„ μŠ€ 및 상업적 λ„μž… μ•ˆλ‚΄)

  • Non-Commercial Use Only (비상업적 μ—°κ΅¬Β·κ°œμΈ 이용 ν•œμ •):
    This model is strictly provided for non-commercial, research, and educational purposes only. Any commercial use without a separate commercial agreement is strictly prohibited.
    λ³Έ λͺ¨λΈμ€ 순수 연ꡬ, ꡐ윑 및 개인 λΉ„μ˜λ¦¬ λͺ©μ μ— ν•œν•΄μ„œλ§Œ 무료둜 μ œκ³΅λ©λ‹ˆλ‹€. 사전 ν—ˆκ°€ μ—†λŠ” 일체의 영리적 μ΄μš©μ€ μ—„κ²©νžˆ κΈˆμ§€λ©λ‹ˆλ‹€.

  • Commercial Licensing & Inquiries for Revenue-Generating Entities (맀좜 λ°œμƒ κΈ°μ—… 및 μƒμš©ν™” λ„μž… 문의):
    If your organization generates revenue, or if you plan to incorporate this model into commercial products, services, or paid APIs, you are required to purchase a commercial license. Please contact us directly via email for enterprise licensing and integration support.
    κΈ°μ—…/사업체에 맀좜이 λ°œμƒν•˜κ³  μžˆκ±°λ‚˜, λ³Έ λͺ¨λΈμ„ μƒμš© μ„œλΉ„μŠ€Β·μœ λ£Œ μ œν’ˆΒ·μˆ˜μ΅ 창좜 λͺ©μ μ˜ μ‹œμŠ€ν…œμ— λ„μž…ν•˜κ³ μž ν•˜μ‹œλŠ” 경우 λ°˜λ“œμ‹œ λ³„λ„μ˜ μƒμš© λΌμ΄μ„ μŠ€λ₯Ό μ·¨λ“ν•˜μ…”μ•Ό ν•©λ‹ˆλ‹€. λ„μž… 쑰건 및 κΈ°μ—…μš© λΌμ΄μ„ μŠ€ λ°œκΈ‰μ€ μ•„λž˜ μ΄λ©”μΌλ‘œ λ¬Έμ˜ν•΄ μ£Όμ‹œκΈ° λ°”λžλ‹ˆλ‹€.

    • λ„μž… 및 μƒμš© λΌμ΄μ„ μŠ€ 문의 (Commercial Inquiries): admin@099.kr
    • 곡식 μ›Ήμ‚¬μ΄νŠΈ (Official Website): https://baar.uk/model/

⚠️ Notes & Disclaimer (기술 κ³ μ§€ 및 μ•ˆλ‚΄)

  • Defensive Engineering & Evaluation: λ³Έ λͺ¨λΈμ€ 제둜 트러슀트 봇 차단 및 이상 ν–‰μœ„ 탐지 연ꡬλ₯Ό μœ„ν•΄ μ œμž‘λ˜μ—ˆμŠ΅λ‹ˆλ‹€. μ‹€μ œ 운영 ν™˜κ²½ 배치 μ‹œ, μžμ‚¬ μ„œλΉ„μŠ€ νŠΈλž˜ν”½ νŠΉμ„±μ— λ§žμΆ”μ–΄ 차단 μž„κ³„κ°’(config.json)을 사전에 μ‘°μ •ν•˜μ—¬ μ μš©ν•˜μ‹­μ‹œμ˜€.

πŸ’Ό Opportunities & Contact (μ±„μš© μ œμ•ˆ 및 νŒŒνŠΈλ„ˆμ‹­)

This project demonstrates end-to-end competency in multimodal behavioral threat defense, bot heuristic spoofing mitigation, edge quantization (INT8/WASM), and ultra-low latency deployment.

I am actively seeking AI Engineering / Security AI Research opportunities, recruitment offers, and technical partnerships.

  • 곡식 μ›Ήμ‚¬μ΄νŠΈ (Official Website): https://baar.uk/model/
  • κΈ°μ—… λ„μž… 및 λΌμ΄μ„ μŠ€ 문의 (Commercial Adoption): admin@099.kr
  • μ±„μš© 및 ν¬μ§€μ…˜ μ œμ•ˆ (Recruitment & Hiring): admin@099.kr
  • 투자 및 기술 ν˜‘λ ₯ 문의 (Investment & Partnerships): admin@099.kr

극단적인 λ¦¬μ†ŒμŠ€ ν™˜κ²½(Edge, CDN, WASM)μ—μ„œλ„ μ‹€μ‹œκ°„ 좔둠이 κ°€λŠ₯ν•œ λͺ¨λΈμ„ μ„€κ³„ν•˜κ³ , μ‹ μ’… 봇/μ—μ΄μ „νŠΈ μœ„ν˜‘ λ°©μ–΄ μ—”μ§„ ꡬ좕뢀터 INT8 μ–‘μžν™” 및 κΈ€λ‘œλ²Œ λ°°ν¬κΉŒμ§€ μ „ 과정을 직접 κ΅¬ν˜„ν•˜λŠ” AI μ—”μ§€λ‹ˆμ–΄μž…λ‹ˆλ‹€. 기술 λ„μž…, μ±„μš© μ œμ•ˆ 및 ν˜‘μ—…μ— 관심 μžˆλŠ” κΈ°μ—…κ³Ό νŒ€μ˜ 문의λ₯Ό ν™˜μ˜ν•©λ‹ˆλ‹€.

Downloads last month
23
Safetensors
Model size
57.4k params
Tensor type
F32
Β·
Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support