Qriton Shield

Real-time IP threat intelligence with Modern Hopfield Networks. Detect and block malicious IPs with AI-powered anomaly detection, a 4-tier response system, and a global network of Shields.

Installation

pip install qriton-shield

Quick Start

from datetime import datetime
from qriton_shield import Shield, HttpRequest

shield = Shield()
shield.start()

# Process an HTTP request
result = shield.process_request(HttpRequest(
    timestamp=datetime.now(),
    client_ip="192.168.1.100",
    method="GET",
    url="/api/data",
    status_code=200,
    bytes_sent=1024,
    bytes_received=256,
    user_agent="Mozilla/5.0",
))

print(f"Decision: {result.decision}")  # allow | rate_limit | challenge | block
print(f"Tier: {result.tier}")          # 0-3
print(f"Threat Score: {result.threat_score}")
print(f"Anomaly Score: {result.anomaly_score}")

shield.stop()

Features

• Modern Hopfield Networks — Continuous Hopfield Network with LogSumExp energy function for exponential-capacity anomaly detection
• 20-Feature Analysis — Network (L4), application (L7), and malware (V2) traffic features with gradient-based attribution
• 4-Tier Response — Allow (0-30) -> Rate Limit (30-60) -> Challenge (60-80) -> Block (80+)
• Adaptive Threat Modes — relaxed / balanced / aggressive / lockdown with auto-escalation
• Web Application Firewall — OWASP Top 10 pattern detection (SQLi, XSS, path traversal, command injection)
• Cross-Platform Firewall — Windows routes, Linux iptables, macOS pfctl
• Subnet Velocity Detection — Auto-aggregate /24 blocks for coordinated attacks
• Challenge/Trust System — Behavioral verification with trust tokens
• Night Mode — Automatic minimum protection during off-hours
• Geographic Fencing — Country-based blocking with risk multipliers
• Threat Intelligence — Spamhaus, Firehol, AbuseIPDB integration
• CDN Whitelist — Cloudflare, AWS CloudFront, Google, Fastly auto-whitelisted
• Federated Learning — Cross-shield Hopfield model sharing
• Data Exfiltration Detection — Credit cards, SSNs, API keys, private keys
• Troll Mode — Rickrolls, tarpits, honeypots, fake vulnerabilities
• MCP Guard — Firewall for Claude MCP tool calls (prompt injection detection)
• Hash-Chain Audit Log — Tamper-evident decision trail for compliance
• 7-Layer Pipeline (v2) — Network, Transport, Identity, Capability, Semantic, State, Collective

CLI

# Start Shield server
shield -m balanced -p 8765

# Start with config file
shield -c settings.json

# Monitor a log file
shield -l /var/log/nginx/access.log

# Security audit
shield audit

# Version
shield version

API

from qriton_shield import Shield, ShieldConfig, NightModeConfig

# Full configuration
config = ShieldConfig(
    threat_mode="balanced",
    state_path="shield_state.json",
    night_mode=NightModeConfig(enabled=True, start_hour=22, end_hour=6, minimum_mode="balanced"),
)

shield = Shield(config)
shield.on("block", lambda d: print(f"Blocked {d['ip']}: {d['reason']}"))
shield.on("anomaly", lambda d: print(f"Anomaly score: {d['score']:.1f}"))
shield.start()

Hopfield Network (standalone)

from qriton_shield import ContinuousHopfieldNetwork

net = ContinuousHopfieldNetwork(20, beta=1.0)
net.train([pattern1, pattern2, pattern3])
result = net.recall(noisy_input)
print(f"Energy: {result.final_energy}, Converged: {result.converged}")

WAF

from qriton_shield import WafEngine

waf = WafEngine()
match = waf.check("/search?q=1' UNION SELECT * FROM users--")
if match:
    print(f"WAF: {match.category} ({match.rule_name}), score={match.score}")

MCP Guard

from qriton_shield import MCPGuard

guard = MCPGuard()
result = guard.check_tool_call("filesystem_write", {"path": "/etc/passwd", "content": "..."})
if not result["allowed"]:
    print(f"Blocked: {result['reason']}")

Geographic Fencing

from qriton_shield import GeoFencing, GeoConfig

geo = GeoFencing(GeoConfig(
    enabled=True,
    whitelist=["US", "DE", "GB"],
    high_risk_countries=["CN", "RU", "KP"],
))
info = geo.check_ip("8.8.8.8")
print(f"Country: {info['country']}, High Risk: {info['is_high_risk']}")

Threat Modes

Architecture

HTTP Request
    |
    v
[Whitelist Check] --> bypass if CDN/known-good
    |
[Threat Scoring] --> L4 + L7 + credential stuffing + rate acceleration
    |
[Hopfield AI] --> 20-feature anomaly detection with explainability
    |
[4-Tier Response]
    |-- Tier 0: Allow (score < 30)
    |-- Tier 1: Rate Limit (30-60)
    |-- Tier 2: Challenge (60-80)
    |-- Tier 3: Block (80+)
    |
[Auto Mode Switch] --> escalate/deescalate based on block rate

Requirements

• Python >= 3.10
• No external dependencies (pure Python)

Links

• Homepage: https://shield.qriton.com
• NPM Package: https://www.npmjs.com/package/@qriton/shield
• Repository: https://github.com/qriton/shield

License

Apache-2.0