Skip to content

Latest commit

 

History

20 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

HookCheck

HookCheck is a free, public, privacy-first decision-support tool that detects phishing and scams instantly. No accounts, no login, no personal data stored.

Submit a suspicious message, URL, screenshot, or QR code. HookCheck aggregates multi-layered security signals and explains its reasoning in clear, plain language.


Detection Architecture

HookCheck runs two independent evidence layers on every submission. No single layer silently overrides the other — every report displays findings separately by source so users can make informed decisions.

1. Deterministic Heuristics Engine (Local, Uninjectable)

The first layer runs entirely server-side without external network calls. Because it is deterministic, it cannot be bypassed or fooled by adversarial input.

Typosquatting & Homoglyphs: Calculates edit distances (Levenshtein ≤ 2) against a curated list of commonly-impersonated domains and flags Unicode substitutions that visually spoof legitimate brands. - Brand Subdomain Impersonation: Flags when trusted brand names appear as subdomains of unrelated registered domains (e.g. paypal.com. login-verify.ru). - TLD & URL Features: Identifies suspicious top-level domains commonly used in phishing, excessive URL length, known shorteners, plaintext HTTP, and credential-harvesting keywords in paths. - Text Pattern Analysis: Detects social engineering language patterns — urgency cues, payment demands, credential harvesting phrases, and suspicious contact channels.

2. VirusTotal Threat Intelligence

Queries VirusTotal's public registry to check domains against 70+ antivirus and threat intelligence engines.

  • Corroborating Evidence: VirusTotal findings are treated as supporting data — a "clean" result does not suppress strong flags raised by heuristics, preventing newly registered phishing campaigns from bypassing detection.

Scoring

The overall threat level is computed from a weighted combination of both layers (60% Heuristics, 40% VirusTotal). Floor rules ensure that strong signals from either layer are never diluted below a meaningful threshold. The result is presented as a categorical verdict (Likely Safe, Medium Risk, High Risk, or Critical Threat) rather than a numeric score.


Features

  • Unified Submission Surface: A single interface that automatically detects input type. Paste text, paste a URL, or drop an image — no manual mode selectors. - QR Code & Screenshot Processing: Decodes QR codes for quishing protection and extracts text from screenshots via OCR, passing payloads through the full analysis pipeline. - Privacy-First Cache: Results are cached for up to 24 hours using only SHA-256 hashes of normalised inputs. Raw submissions are never persisted. - Verdict-First Interface: A clear threat badge and plain-language summary up front, followed by granular per-layer evidence panels. - Shareable Reports: Analysis reports are available via non- sequential, randomly generated permalink URLs. All report pages are marked noindex to prevent search engine indexing.

About

A free phishing and scam detector that analyzes text, links, screenshots, and QR codes using domain heuristics and VirusTotal threat intelligence.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Contributors

Languages