View Issue Details

Jump to NotesJump to History
IDProjectCategoryView StatusDate SubmittedLast Update
0009641Kali LinuxNew Tool Requestspublic2026-04-16 19:332026-04-16 19:33
ReporterNuClide Assigned To 
PrioritynormalSeverityminorReproducibilityhave not tried
Status newResolutionopen 
Summary0009641: [New Tool] aimap - AI infrastructure scanner (LLMs, vector DBs, ML servers)
Description

Name of the tool

aimap

Version

v1.1.1

Homepage

https://github.com/Nicholas-Kloster/aimap

Download URL

https://github.com/Nicholas-Kloster/aimap/releases/latest

Description

aimap is a purpose-built scanner for AI and machine learning infrastructure.
Where nmap and nuclei identify that a service is running, aimap identifies
which AI/ML service is running and whether it is exposing data, PII,
credentials, or compute.

It fingerprints 23 AI/ML service types across six categories (vector
databases, LLM runtimes, ML platforms, orchestration/UI, observability,
notebooks) and runs service-specific deep enumeration to surface actionable
findings:

  • Unauthenticated Jupyter kernels (RCE)

  • Vector database PII field exposure

  • Flowise credentials endpoint accessibility

  • Ollama open inference endpoints (LLMjacking vector)

  • Dify with unclaimed admin accounts

  • Langfuse instances leaking LLM conversation history

  • MLflow experiments and model registries accessible

  • Weaviate schemas with regulated-data fields

  • ChromaDB collections with PII-indicating names

    Fingerprinted services: Weaviate, ChromaDB, Qdrant, Milvus, Ollama, vLLM,
    LocalAI, text-generation-webui, MLflow, TensorFlow Serving, Triton
    Inference Server, Ray Serve, Ray Dashboard, Kubeflow, LangServe, Flowise,
    Dify, Open WebUI, LiteLLM, BentoML, Langfuse, Jupyter Notebook, Docker
    Registry.

    Dependencies

    None. Single statically-linked Go binary. Built with Go stdlib only —
    no external Go modules, no system libraries, no runtime dependencies.

    License

    MIT

    Similar tools

    None in Kali currently. Generic scanners (nmap, nuclei) can detect HTTP
    services on AI-related ports but do not fingerprint them as AI services
    specifically, do not perform service-appropriate deep enumeration, and do
    not surface findings like "ChromaDB collection with PII field exposed" or
    "Jupyter with unauthenticated code execution."

    aimap fills a gap that has become significant as AI adoption in
    enterprise environments has accelerated. Security teams auditing their
    own networks for shadow AI deployments — or bug bounty researchers
    testing AI companies within program scope — currently have to chain
    together multiple tools and hand-craft detections. aimap provides this
    coverage natively.

    Why this should be in Kali

    AI infrastructure exposure has become a mainstream attack surface:

  • Operation Bizarre Bazaar (disclosed January 2026) documented a
    commercial marketplace built on compromised LLM inference servers
    across 30+ providers.

  • Pillar Security, Cisco Talos, and UpGuard have all published research
    on unauthenticated vector database exposure during 2024-2025.

  • GreyNoise has observed 91,000+ documented attack sessions targeting
    AI/ML services on Shodan-indexed infrastructure.

  • The OWASP LLM Top 10 now formalizes several of these exposure classes.

    Kali currently lacks any tool specifically designed for AI
    infrastructure reconnaissance. aimap is a natural addition alongside
    nmap, nuclei, and masscan in the /auditing category.

    Usage examples

    Shadow-AI audit of internal network

    aimap -target 10.0.0.0/24 -threads 50 -o audit.json

    Single-host deep dive with wide port coverage

    aimap -target 10.5.5.5 -v -ports 8000,8080,8888,9091,11434,6333,19530

    CI/CD deployment gate

    aimap -target $DEPLOY_URL -o check.json && \
    jq '.enum_results[] | select(.risk_level == "critical")' check.json

    External exposure check against your own public ranges

    aimap -list corp-public-ips.txt -threads 10 -timeout 10s -o external.json

    Technical details

  • Single Go binary (~5MB statically linked, prebuilt for linux+darwin amd64+arm64)

  • Read-only HTTP GETs only (no writes, no auth attempts, no exploits)

  • 3-phase pipeline: port scan → service fingerprint → deep enumeration

  • JSON output with stable schema for pipeline integration

  • Colored terminal output with risk scoring

  • CIDR support, threaded scanning, configurable timeouts

  • Man page included (aimap.1)

  • BlackArch PR pending: https://github.com/BlackArch/blackarch/pull/4912

    Build

    go install github.com/Nicholas-Kloster/aimap@latest

    Or from source:

    git clone https://github.com/Nicholas-Kloster/aimap.git
    cd aimap
    go build -o aimap .

    Or download a prebuilt binary:

    curl -LO https://github.com/Nicholas-Kloster/aimap/releases/download/v1.1.1/aimap-linux-amd64
    chmod +x aimap-linux-amd64
    sudo mv aimap-linux-amd64 /usr/local/bin/aimap

    Tested on: Go 1.21+ / Ubuntu 22.04, 24.04 / Arch / Kali Linux rolling

    Author / contact

    Nicholas Kloster (NuClide)
    GitHub: https://github.com/Nicholas-Kloster
    Issues: https://github.com/Nicholas-Kloster/aimap/issues

Attached Files
PKGBUILD (1,226 bytes)    
# Maintainer: Nicholas Kloster <[email protected]>

pkgname=aimap
pkgver=1.1.1
pkgrel=1
groups=('blackarch' 'blackarch-scanner' 'blackarch-recon' 'blackarch-networking')
pkgdesc='Security scanner for AI and ML infrastructure. Fingerprints 23 AI/ML service types (LLMs, vector databases, model servers) and surfaces actionable findings.'
arch=('x86_64' 'aarch64')
url='https://github.com/Nicholas-Kloster/aimap'
license=('MIT')
makedepends=('go')
source=("$pkgname-$pkgver.tar.gz::https://github.com/Nicholas-Kloster/aimap/archive/v$pkgver.tar.gz")
sha256sums=('c1a48b61524f1f3d4c5e046c1a84189db36e55bce1900ede1107cfbabbb1246c')

build() {
  cd "$pkgname-$pkgver"
  export CGO_ENABLED=0
  export GOFLAGS="-trimpath -mod=readonly -modcacherw"
  export LDFLAGS="-buildmode=pie -linkmode=external -s -w"
  go build -o "$pkgname" .
}

package() {
  cd "$pkgname-$pkgver"

  # Binary
  install -Dm755 "$pkgname" "$pkgdir/usr/bin/$pkgname"

  # Man page
  if [ -f "aimap.1" ]; then
    install -Dm644 "aimap.1" "$pkgdir/usr/share/man/man1/aimap.1"
  fi

  # License
  install -Dm644 "LICENSE" "$pkgdir/usr/share/licenses/$pkgname/LICENSE"

  # Documentation
  install -Dm644 "README.md" "$pkgdir/usr/share/doc/$pkgname/README.md"
}
PKGBUILD (1,226 bytes)   

Activities

Issue History

Date Modified Username Field Change
2026-04-16 19:33 NuClide New Issue
2026-04-16 19:33 NuClide File Added: PKGBUILD