cpex-url-reputation 0.3.0

High-performance URL reputation and phishing detection for MCP Gateway

URL Reputation (Rust)

Author: Matheus Cafalchio Version: 0.1.0

Blocks URLs based on configured blocked domains, patterns and heuristics before resource fetch. Designed for fast and efficient resource checks.

Runtime Requirements

This plugin depends on cpex>=0.1.0rc1,<0.2 and imports hook models from cpex.framework. The compiled Rust extension is mandatory; there is no Python fallback implementation.

Hooks

• resource_pre_fetch – triggered before any resource is fetched.

Config

config:
    whitelist_domains: ["ibm.com", "yourdomain.com"]
    allowed_patterns: ["^https://trusted\\.internal/.*"]
    blocked_domains: ["malicious.example.com"]
    blocked_patterns: ["casino", "crypto"]
    use_heuristic_check: true
    entropy_threshold: 3.65
    block_non_secure_http: true

Config Description

• whitelist_domains

  • A set of domains that are allowed to be fetched without any checks.

• allowed_patterns

  • A list of regex patterns matched against the full URL. If any pattern matches, the URL is allowed and skips all remaining checks — including the non-secure HTTP check. Evaluated after the whitelist, before scheme enforcement.

• blocked_domains

  • A set of domains that will always be blocked.

• blocked_patterns

  • A list of regex patterns matched against the full URL. If any pattern matches, the URL is blocked.

• use_heuristic_check

  • Whether heuristic checks (entropy, TLD validity, unicode security) should be performed. Default: false.

• entropy_threshold

  • Maximum allowed Shannon entropy for a domain. Higher entropy may indicate suspicious/malicious domains.

• block_non_secure_http

  • Whether URLs using http (non-secure) should be blocked. Default: true.

Architecture

flowchart LR
    Start([URL Input]) --> Parse{Parse & Extract Domain}
    Parse -->|Fail| Block1[❌ Parse Error]
    Parse -->|Success| DetectIP[Detect IP]

    DetectIP --> Whitelist{Whitelist?}
    Whitelist -->|Yes| Success[✅ Allow]
    Whitelist -->|No| AllowPat{Allowed Pattern?}

    AllowPat -->|Yes| Success
    AllowPat -->|No| HTTP{Scheme = HTTPS<br/>or not enforced?}

    HTTP -->|No| Block2[❌ Non-HTTPS]
    HTTP -->|Yes| BlockedDom{Blocked Domain<br/>or Pattern?}

    BlockedDom -->|Yes| Block3[❌ Blocked]
    BlockedDom -->|No| Heuristic{Heuristic Check<br/>Enabled & Not IP?}

    Heuristic -->|No| Success
    Heuristic -->|Yes| Checks{Pass Entropy,<br/>TLD & Unicode?}

    Checks -->|No| Block4[❌ Heuristic Fail]
    Checks -->|Yes| Success

    Block1 --> End([Return])
    Block2 --> End
    Block3 --> End
    Block4 --> End
    Success --> End

    style Start fill:#e1f5ff
    style End fill:#e1f5ff
    style Success fill:#c8e6c9
    style Block1 fill:#ffcdd2
    style Block2 fill:#ffcdd2
    style Block3 fill:#ffcdd2
    style Block4 fill:#ffcdd2

Logic workflow

1. Parse & Normalize URL

   • Trim the input URL, then parse it (scheme and host are normalised to lowercase by the URL parser per RFC 3986; path and query retain original casing).
   • Fail → Violation: "Could not parse url".

2. Extract Domain

   • Get the host string from the URL.
   • Fail → Violation: "Could not parse domain".

3. Detect IP Address

   • Determine if domain is an IPv4 or IPv6 address.
   • Skip heuristic checks for IPs.

4. Whitelist Check

   • If domain is in whitelist_domains → continue_processing = true, skip all further checks.

5. Allowed Patterns Check

   • If URL matches any regex in allowed_patterns → continue_processing = true, skip all further checks.
   • Note: this check runs before scheme enforcement, so an allowed_patterns match can bypass the non-secure HTTP block.

6. Block Non-Secure HTTP

   • If scheme ≠ "https" and block_non_secure_http → Violation: "Blocked non secure http url".

7. Blocked Domains

   • If domain is in blocked_domains → Violation: "Domain in blocked set".

8. Blocked Patterns

   • If URL matches any regex in blocked_patterns → Violation: "Blocked pattern".

9. Heuristic Checks (only for non-IP domains and if use_heuristic_check = true): 9.1 High Entropy Check – If Shannon entropy > entropy_threshold → Violation: "High entropy domain". 9.2 TLD Validity Check – Validate top-level domain. Fail → Violation: "Illegal TLD". 9.3 Unicode Security Check – Validate domain unicode. Fail → Violation: "Domain unicode is not secure".

10. Final Outcome

    • If no violations → continue_processing = true.
    • If any check fails → return first PluginViolation and continue_processing = false.

Limitations

- Static lists only; no external reputation providers.
- Ianna valid TLDs are static and will be out of date
- Ignores other schemes that are not http and https
- No external domain reputation checks

TODOs

- External threat-intel integration with cache – Query external feeds for known malicious domains.
- IP address handling policy – Decide rules for IPv4/IPv6 URLs.
- Dynamic TLD updates – Fetch latest IANA TLD list automatically.

Tests

Test Coverage (24 unit tests, all passing):

Filename	Function Coverage	Line Coverage	Region Coverage
engine.rs	96.55% (28/29)	99.26% (533/537)	98.60% (634/643)
filters/heuristic.rs	100.00% (5/5)	96.49% (55/57)	97.53% (79/81)
filters/patterns.rs	100.00% (5/5)	100.00% (20/20)	100.00% (38/38)
lib.rs	0.00% (0/1)	0.00% (0/5)	0.00% (0/7)
types.rs	50.00% (3/6)	44.12% (15/34)	23.94% (17/71)
TOTAL	89.13% (41/46)	95.43% (627/657)	91.45% (770/842)

Note: lib.rs and types.rs contain PyO3 bindings and module declarations not covered by unit tests.

New test coverage includes:

• Invalid regex pattern handling (both allowed and blocked patterns)
• Case-insensitive domain matching (whitelist and blocklist)
• Subdomain matching validation

Run tests:

cargo nextest run -p url_reputation  # Run Rust unit tests
cargo llvm-cov --lib --html   # Generate coverage report

Heuristic methods

The heuristics were based on a research paper.

A. P. S. Bhadauria and M. Singh, "Domain‑Checker: A Classification of Malicious and Benign Domains Using Multitier Filtering," Springer Nature, 2023.