performing-agentless-vulnerability-scanning

Imported from mukul975/anthropic-cybersecurity-skills (skills/performing-agentless-vulnerability-scanning/SKILL.md). Install upstream with npx skills add mukul975/anthropic-cybersecurity-skills --skill performing-agentless-vulnerability-scanning. Copyright stays with the author (Apache-2.0).

Performing Agentless Vulnerability Scanning

Overview

Agentless vulnerability scanning assesses systems for security weaknesses without requiring endpoint agent installation. This approach leverages existing network protocols (SSH for Linux, WMI for Windows), cloud provider APIs for snapshot-based analysis, and authenticated remote checks. Modern cloud platforms like Microsoft Defender for Cloud, Wiz, Datadog, and Tenable perform out-of-band analysis by taking disk snapshots and examining OS configurations and installed packages offline. The open-source tool Vuls provides agentless scanning based on NVD and OVAL data for Linux/FreeBSD systems. This skill covers configuring agentless scans across on-premises, cloud, and containerized environments.

When to Use

• When conducting security assessments that involve performing agentless vulnerability scanning
• When following incident response procedures for related security events
• When performing scheduled security testing or auditing activities
• When validating security controls through hands-on testing

Prerequisites

• SSH key-based authentication configured on Linux/Unix targets
• WMI/WinRM access on Windows targets with appropriate credentials
• Cloud provider API credentials (AWS IAM, Azure RBAC, GCP IAM)
• Network access from scanner to target systems on required ports
• Service account with read-only access to target system configurations
• Python 3.8+ for custom scanning automation

Core Concepts

Agentless vs Agent-Based Scanning

Agentless Scanning Methods

Cloud Snapshot Analysis Flow

1. Scanner requests disk snapshot via cloud API
2. Cloud provider creates snapshot of VM root + data disks
3. Scanner mounts snapshot in isolated analysis environment
4. Scanner examines OS packages, configurations, file system
5. Snapshot is deleted after analysis (no persistent copies)
6. Results sent to central management console

Workflow

Step 1: SSH-Based Agentless Scanning (Linux)

# Create dedicated scan SSH key pair
ssh-keygen -t ed25519 -f /opt/scanner/.ssh/scan_key -N "" \
  -C "vuln-scanner@security.local"

# Deploy public key to targets via Ansible
# ansible-playbook deploy_scan_key.yml

# Test connectivity to target
ssh -i /opt/scanner/.ssh/scan_key -o ConnectTimeout=10 \
  scanner@target-host "cat /etc/os-release && dpkg -l 2>/dev/null || rpm -qa"

import paramiko
import json

class AgentlessLinuxScanner:
    """SSH-based agentless vulnerability scanner for Linux systems."""

    def __init__(self, key_path):
        self.key_path = key_path

    def connect(self, hostname, username="scanner", port=22):
        """Establish SSH connection to target."""
        client = paramiko.SSHClient()
        client.set_missing_host_key_policy(paramiko.AutoAddPolicy())
        key = paramiko.Ed25519Key.from_private_key_file(self.key_path)
        client.connect(hostname, port=port, username=username, pkey=key,
                       timeout=30, banner_timeout=30)
        return client

    def get_os_info(self, client):
        """Detect OS type and version."""
        _, stdout, _ = client.exec_command("cat /etc/os-release", timeout=10)
        os_release = stdout.read().decode()
        info = {}
        for line in os_release.strip().split("\n"):
            if "=" in line:
                key, val = line.split("=", 1)
                info[key] = val.strip('"')
        return info

    def get_installed_packages(self, client):
        """Enumerate installed packages."""
        # Try dpkg (Debian/Ubuntu)
        _, stdout, _ = client.exec_command(
            "dpkg-query -W -f='${Package}|${Version}|${Architecture}\\n'",
            timeout=30
        )
        output = stdout.read().decode().strip()
        if output:
            packages = []
            for line in output.split("\n"):
                parts = line.split("|")
                if len(parts) >= 2:
                    packages.append({
                        "name": parts[0],
                        "version": parts[1],
                        "arch": parts[2] if len(parts) > 2 else "",
                        "manager": "dpkg"
                    })
            return packages

        # Try rpm (RHEL/CentOS/Fedora)
        _, stdout, _ = client.exec_command(
            "rpm -qa --queryformat '%{NAME}|%{VERSION}-%{RELEASE}|%{ARCH}\\n'",
            timeout=30
        )
        output = stdout.read().decode().strip()
        packages = []
        for line in output.split("\n"):
            parts = line.split("|")
            if len(parts) >= 2:
                packages.append({
                    "name": parts[0],
                    "version": parts[1],
                    "arch": parts[2] if len(parts) > 2 else "",
                    "manager": "rpm"
                })
        return packages

    def check_kernel_version(self, client):
        """Get running kernel version."""
        _, stdout, _ = client.exec_command("uname -r", timeout=10)
        return stdout.read().decode().strip()

    def check_listening_ports(self, client):
        """Enumerate listening network services."""
        _, stdout, _ = client.exec_command(
            "ss -tlnp 2>/dev/null || netstat -tlnp 2>/dev/null",
            timeout=10
        )
        return stdout.read().decode().strip()

    def scan_host(self, hostname, username="scanner"):
        """Perform full agentless scan of a host."""
        print(f"[*] Scanning {hostname}...")
        client = self.connect(hostname, username)

        result = {
            "hostname": hostname,
            "os_info": self.get_os_info(client),
            "kernel": self.check_kernel_version(client),
            "packages": self.get_installed_packages(client),
            "listening_ports": self.check_listening_ports(client),
        }

        client.close()
        print(f"  [+] Found {len(result['packages'])} packages on {hostname}")
        return result

Step 2: WinRM-Based Agentless Scanning (Windows)

import winrm

class AgentlessWindowsScanner:
    """WinRM-based agentless vulnerability scanner for Windows."""

    def __init__(self, username, password, domain=None):
        self.username = username
        self.password = password
        self.domain = domain

    def connect(self, hostname, use_ssl=True):
        """Create WinRM session."""
        port = 5986 if use_ssl else 5985
        transport = "ntlm"
        user = f"{self.domain}\\{self.username}" if self.domain else self.username
        session = winrm.Session(
            f"{'https' if use_ssl else 'http'}://{hostname}:{port}/wsman",
            auth=(user, self.password),
            transport=transport,
            server_cert_validation="ignore"
        )
        return session

    def get_installed_hotfixes(self, session):
        """Get installed Windows updates/hotfixes."""
        cmd = "Get-HotFix | Select-Object HotFixID,InstalledOn,Description | ConvertTo-Json"
        result = session.run_ps(cmd)
        if result.status_code == 0:
            return json.loads(result.std_out.decode())
        return []

    def get_installed_software(self, session):
        """Enumerate installed software from registry."""
        cmd = """
        $paths = @(
            'HKLM:\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\Uninstall\\*',
            'HKLM:\\SOFTWARE\\WOW6432Node\\Microsoft\\Windows\\CurrentVersion\\Uninstall\\*'
        )
        Get-ItemProperty $paths -ErrorAction SilentlyContinue |
            Where-Object {$_.DisplayName} |
            Select-Object DisplayName, DisplayVersion, Publisher |
            ConvertTo-Json
        """
        result = session.run_ps(cmd)
        if result.status_code == 0:
            return json.loads(result.std_out.decode())
        return []

    def get_os_info(self, session):
        """Get Windows OS details."""
        cmd = "Get-CimInstance Win32_OperatingSystem | Select-Object Caption,Version,BuildNumber,OSArchitecture | ConvertTo-Json"
        result = session.run_ps(cmd)
        if result.status_code == 0:
            return json.loads(result.std_out.decode())
        return {}

    def scan_host(self, hostname):
        """Perform full agentless scan of Windows host."""
        print(f"[*] Scanning {hostname} via WinRM...")
        session = self.connect(hostname)

        result = {
            "hostname": hostname,
            "os_info": self.get_os_info(session),
            "hotfixes": self.get_installed_hotfixes(session),
            "software": self.get_installed_software(session),
        }

        print(f"  [+] Found {len(result['hotfixes'])} hotfixes, "
              f"{len(result['software'])} software entries")
        return result

Step 3: Cloud Snapshot Scanning (AWS)

import boto3
import time

class AWSSnapshotScanner:
    """AWS EC2 agentless snapshot-based vulnerability scanner."""

    def __init__(self, region="us-east-1"):
        self.ec2 = boto3.client("ec2", region_name=region)

    def create_snapshot(self, volume_id, description="Security scan snapshot"):
        """Create EBS snapshot for analysis."""
        snapshot = self.ec2.create_snapshot(
            VolumeId=volume_id,
            Description=description,
            TagSpecifications=[{
                "ResourceType": "snapshot",
                "Tags": [
                    {"Key": "Purpose", "Value": "VulnScan"},
                    {"Key": "AutoDelete", "Value": "true"},
                ]
            }]
        )
        snapshot_id = snapshot["SnapshotId"]
        print(f"  [*] Creating snapshot {snapshot_id} from {volume_id}...")

        waiter = self.ec2.get_waiter("snapshot_completed")
        waiter.wait(SnapshotIds=[snapshot_id])
        print(f"  [+] Snapshot {snapshot_id} ready")
        return snapshot_id

    def delete_snapshot(self, snapshot_id):
        """Clean up snapshot after analysis."""
        self.ec2.delete_snapshot(SnapshotId=snapshot_id)
        print(f"  [+] Deleted snapshot {snapshot_id}")

    def scan_instance(self, instance_id):
        """Scan an EC2 instance via snapshot analysis."""
        print(f"[*] Agentless scan of instance {instance_id}")

        instance = self.ec2.describe_instances(
            InstanceIds=[instance_id]
        )["Reservations"][0]["Instances"][0]

        root_volume = None
        for bdm in instance.get("BlockDeviceMappings", []):
            if bdm["DeviceName"] == instance.get("RootDeviceName"):
                root_volume = bdm["Ebs"]["VolumeId"]
                break

        if not root_volume:
            print("  [!] No root volume found")
            return None

        snapshot_id = self.create_snapshot(root_volume)
        try:
            # Analysis would be performed here
            # Mount snapshot, examine packages, check configs
            result = {
                "instance_id": instance_id,
                "snapshot_id": snapshot_id,
                "root_volume": root_volume,
                "platform": instance.get("Platform", "linux"),
                "state": instance["State"]["Name"],
            }
            return result
        finally:
            self.delete_snapshot(snapshot_id)

Step 4: Vuls Open-Source Agentless Scanner

# /etc/vuls/config.toml - Vuls configuration for agentless scanning

[servers]

[servers.web-server-01]
host = "192.168.1.10"
port = "22"
user = "vuls"
keyPath = "/opt/vuls/.ssh/scan_key"
scanMode = ["fast"]

[servers.db-server-01]
host = "192.168.1.20"
port = "22"
user = "vuls"
keyPath = "/opt/vuls/.ssh/scan_key"
scanMode = ["fast-root"]
[servers.db-server-01.optional]
  [servers.db-server-01.optional.sudo]
    password = ""

[servers.container-host-01]
host = "192.168.1.30"
port = "22"
user = "vuls"
keyPath = "/opt/vuls/.ssh/scan_key"
scanMode = ["fast"]
containersIncluded = ["${running}"]

# Run Vuls agentless scan
vuls scan

# Generate report
vuls report -format-json -to-localfile

# View results
vuls tui

Best Practices

1. Use SSH key-based authentication instead of passwords for Linux scanning
2. Create dedicated service accounts with minimal read-only privileges for scanning
3. Always clean up cloud snapshots after analysis to avoid storage costs and data exposure
4. Combine agentless scanning with agent-based for comprehensive coverage
5. Schedule scans during low-activity periods to minimize any performance impact
6. Rotate scanning credentials regularly and store in a secrets vault
7. Test scanner connectivity before scheduling production scans
8. Use SNMPv3 with authentication and encryption for network device scanning

Common Pitfalls

• Using shared credentials across multiple environments without proper segmentation
• Not cleaning up temporary snapshots in cloud environments
• Assuming agentless scanning has zero performance impact (network and CPU are used)
• Missing WinRM/SSH firewall rules causing scan failures on new deployments
• Not accounting for SSH host key changes causing authentication failures
• Scanning OT/ICS devices with protocols they cannot safely handle

Related Skills

• implementing-rapid7-insightvm-for-scanning
• implementing-wazuh-for-vulnerability-detection
• deploying-osquery-for-endpoint-monitoring
• performing-remediation-validation-scanning