Open Source Vulnerability Scanners by Attack Surface
Open source vulnerability scanners sorted by what they scan, with licenses checked in November 2026 and what each scan looks like from the SOC side.
15 min read

On this page
- How we picked and checked these scanners
- Which scanner for which attack surface
- Licenses and latest releases (checked November 5, 2026)
- Network and host scanning
- OpenVAS (Greenbone Community Edition)
- Nmap with the vulners script
- Web application scanning (DAST)
- ZAP
- Nuclei
- Nikto
- Containers, IaC and dependencies
- Trivy
- Grype
- osv-scanner
- When the scanner gets compromised: the Trivy case
- Endpoints and configuration baselines
- Wazuh vulnerability detection
- OpenSCAP
- What a vulnerability scan looks like from the SOC
- Open source vs free: where Nessus Essentials fits
- Where to practise
The right open source vulnerability scanner depends on what you are scanning. For hosts and network services, use OpenVAS (Greenbone Community Edition). For web applications, ZAP to crawl and Nuclei for known CVEs. For container images and IaC, Trivy or Grype. For code dependencies, osv-scanner. For agent-based checks, Wazuh, and for configuration baselines, OpenSCAP.
A network scan never opens a container image, and an image scanner never sends a packet to your file server, so ranking these tools against each other makes little sense. This guide sorts them by attack surface instead. In a SOC you usually meet a vulnerability scanner as a burst of alerts before you ever run one, so the guide also shows what each scan looks like in your logs.
How we picked and checked these scanners
A scanner made the list if it is maintained (a release in 2026), covers one attack surface well, and a tier 1 or tier 2 analyst can run it in a lab. For each one we checked three things on November 5, 2026:
- Latest release, from the project's GitHub releases page or official download page.
- License, from the LICENSE or COPYING file in the repository, read at the release tag where it mattered. "Open source" below means an OSI-style license; when a project's license is something else, the table says so.
- Commands and defaults, from the README, official docs or source code. That is how we found that the User-Agent advice in older guides no longer holds.
Which scanner for which attack surface
| What you need to scan | Start with | Also useful | Blind spot |
|---|---|---|---|
| Hosts and network services | OpenVAS | Nmap + vulners script for a quick version-to-CVE pass | Unauthenticated scans only see what a service exposes over the network |
| Web applications (DAST) | ZAP | Nuclei for known CVEs and misconfigurations, Nikto for web server checks | Logic flaws, and anything behind a login you didn't configure |
| Container images and IaC | Trivy | Grype for images, directories and SBOMs | What the container does at runtime |
| Code dependencies (SCA) | osv-scanner | Trivy fs, Grype, Dependency-Check | Bugs in code you wrote |
| Endpoints, via an agent | Wazuh vulnerability detection | Assets with no agent installed | |
| Configuration baselines | OpenSCAP | CVEs that aren't part of the profile you evaluate |
Licenses and latest releases (checked November 5, 2026)
| Scanner | Surface | Latest release | License |
|---|---|---|---|
| OpenVAS scanner | Network | v23.52.8 (Oct 29, 2026) | GPL-2.0. The manager, gvmd, is AGPL-3.0 (v26.43.0) |
| Nmap | Network | 7.993 (Oct 23, 2026) | Nmap Public Source License 0.95, based on GPLv2 with extra terms; not OSI-certified |
| ZAP | Web | 2.17.0 (Dec 15, 2025), plus weekly builds | Apache-2.0 |
| Nuclei | Web, network | v3.11.1 (Aug 8, 2026) | MIT |
| Nikto | Web server | 2.6.1 (Jul 31, 2026) | GPL-3.0 for the code; the databases are not GPL |
| Trivy | Images, IaC, dependencies | v0.76.1 (Oct 21, 2026) | Apache-2.0 |
| Grype | Images, dependencies | v0.121.1 (Nov 2, 2026) | Apache-2.0 |
| osv-scanner | Dependencies | v2.7.0 (Oct 12, 2026) | Apache-2.0 |
| Dependency-Check | Dependencies | v13.0.0 (Aug 3, 2026) | Apache-2.0 |
| Wazuh | Endpoints (agent) | v4.14.9 (Oct 15, 2026) | GPLv2 (see note below) |
| OpenSCAP | Configuration | 1.4.6 (Oct 22, 2026) | LGPL-2.1 |
Two license notes. Nmap's own page says the NPSL is meant to comply with the Open Source Definition but has not been through OSI certification, and it forbids shipping Nmap inside proprietary products without an OEM license. For lab and internal use, neither point matters. Wazuh's v4.14.9 LICENSE is GPLv2. On the main branch, the LICENSE now puts the engine under AGPLv3 and makes some shared libraries dual GPLv2 or AGPLv3; no release ships that change yet.
Network and host scanning
OpenVAS (Greenbone Community Edition)
OpenVAS is the closest open source match to a commercial network vulnerability scanner. It sends tests from a feed of vulnerability checks against every host in a target range, and with credentials it logs in and runs local checks against installed packages. Greenbone's manual calls these "local security checks": SSH credentials for Linux and Unix, SMB credentials (Kerberos or NTLM) for Windows.
The quickest lab install is the Greenbone Community Containers with Docker Compose. The guide asks for 2 CPU cores, 4 GB of RAM and 20 GB of disk at minimum, and recommends 4 cores, 8 GB and 60 GB. It also says plainly that it is meant for testing and getting familiar with the tool, not for production setups.
The free feed is the limit. Greenbone's feed comparison lists what the free Community Feed covers and what it leaves out:
| Coverage | Community Feed (free) | Enterprise Feed (paid) |
|---|---|---|
| Widely used software and operating systems (e.g. Microsoft Office, Ubuntu) | Yes | Yes |
| BSI IT-Grundschutz checks | Yes | Yes |
| Enterprise software and network gear (e.g. Cisco, Microsoft Exchange, Palo Alto Networks) | No | Yes |
| CIS Benchmark compliance checks | No | Yes |
| EPSS-based prioritization | No | Yes |
Use it when you need a real network scan of a lab or a small environment. Skip it when your findings have to cover firewalls, VPN appliances or Exchange servers. A clean Community Feed scan of those devices only means nobody tested them.
Nmap with the vulners script
Nmap is a port scanner, and its scripting engine adds a version-to-CVE lookup on top. The vulners script takes the CPEs from version detection, asks the vulners.com API about them, and prints the known CVEs with CVSS scores:
nmap -sV --script vulners --script-args mincvss=7.0 192.168.56.0/24Nmap classes the script as vuln, safe and external. External means the scanning host makes outbound requests to vulners.com with the software names and versions it found. The script tests nothing, so a Linux package with a backported fix still shows the upstream CVEs for its version string. Use the output as a list of leads, and the CVSS score as a sorting key, not a verdict. If the port numbers in the output don't mean much yet, the common ports cheat sheet covers the ones you will see most.
Web application scanning (DAST)
ZAP
ZAP is the general-purpose web scanner: a proxy, a spider and an active scanner in one, now branded "ZAP by Checkmarx" on its site. The stable line is 2.17.0, and the project builds weekly releases from the main branch, typically every Monday.
For a first run, use the baseline scan. It spiders the target for one minute by default, waits for passive scanning to finish and, in the docs' words, "doesn't perform any actual 'attacks'":
docker run -t ghcr.io/zaproxy/zaproxy:stable zap-baseline.py -t https://www.example.comThat makes it safe to wire into CI. The full active scan does send attack payloads, so only point it at systems you own or are authorized to test.
Nuclei
Nuclei runs YAML templates against a target, one template per known issue: a CVE, an exposed admin panel, a default credential. It is fast and focused, and it finds only what a template describes. Version 3.11.1 is MIT-licensed.
nuclei -target example.com -t http/cves/Two defaults from the README show up in your logs. The rate limit is 150 requests per second unless you set -rl. And for out-of-band (OAST) checks, Nuclei uses interactsh: the payload makes the target call back to one of the default out-of-band servers, oast.pro, oast.live, oast.site, oast.online, oast.fun or oast.me. The -ni flag turns that off.
Nikto
Nikto is a web server scanner. It walks a list of known risky paths (its db_tests file) and flags server versions listed in db_outdated. It is old, loud and still maintained (2.6.1, July 2026). The license is split: per the README, the code is GPL-3.0 but the database files "are NOT licensed under the GPL and may only be distributed as part of the official Nikto package". If you plan to reuse its check database in another tool, that line rules it out.
Containers, IaC and dependencies
Trivy
Trivy scans container images, filesystems, remote Git repositories, VM images and Kubernetes, and looks there for vulnerable packages, IaC misconfigurations, secrets and licenses (README):
trivy image python:3.4-alpine
trivy fs --scanners vuln,secret,misconfig myproject/It downloads its vulnerability database as an OCI artifact from mirror.gcr.io/aquasec/trivy-db, falling back to ghcr.io/aquasecurity/trivy-db (CLI reference). Allow those in your CI egress rules or mirror them.
Grype
Grype does the vulnerability half of Trivy's job for images, directories and SBOMs, including SBOMs made with Anchore's Syft (README):
grype alpine:latest
grype sbom:./sbom.jsonIts database comes from grype.anchore.io. Running Grype next to Trivy in a pipeline is a cheap second opinion, because the two use different vulnerability databases and matching logic.
osv-scanner
osv-scanner is Google's front end to the OSV.dev database, built for lockfiles and source trees:
osv-scanner scan source -r ./my-repo
osv-scanner scan image my-image:tagIts README states what leaves your network: package names, versions, ecosystems and file hashes go to the OSV.dev API. Use --offline with a downloaded database when that is not acceptable. If you work in Java or .NET shops, Dependency-Check is the other common choice; it moved to its own GitHub organization (the old repository is archived, the project is not) and wants an NVD API key.
When the scanner gets compromised: the Trivy case
Aqua's advisory GHSA-69fq-xp46-6x23 (CVE-2026-33634) describes the incident. On March 19, 2026, an attacker used compromised credentials to publish a malicious Trivy v0.69.4, force-push 76 of 77 version tags in aquasecurity/trivy-action and replace all 7 tags in aquasecurity/setup-trivy. On March 22 came malicious Docker Hub images tagged 0.69.5 and 0.69.6. The advisory traces the root cause to an attack that began in late February, where credential rotation "was not atomic".
The payload ran before the real scan. It read the GitHub runner's process memory through /proc/<pid>/mem, swept more than 50 paths for SSH keys, cloud credentials, Kubernetes tokens and .env files, encrypted the haul and sent it out. If that failed and a token was available, it created a public tpcp-docs repository on the victim's GitHub account and uploaded the data as a release asset.
What to hunt for, straight from the advisory:
- DNS or proxy hits for
scan.aquasecurtiy.org(note the misspelling) or connections to45.148.10.212. - Repositories named with a
tpcp-docsprefix in your GitHub organization. - Workflow runs between March 19 and 23, 2026 that used trivy-action before 0.35.0, setup-trivy without SHA pinning, or the Trivy images listed in the advisory. Any secret those pipelines could read counts as exposed.
Any scanner in CI runs with the pipeline's secrets, and a mutable tag like @v0.34.0 is a pointer someone else can move. Pin actions to a full commit SHA, as the advisory recommends, and verify release binaries with cosign. The same pattern hit tj-actions/changed-files in March 2025 (CVE-2025-30066), and the GitHub Actions pipeline hijack operation lets you work that case on a SIEM, XDR and Firewall console. ATT&CK files the technique under supply chain compromise.
Endpoints and configuration baselines
Wazuh vulnerability detection
Wazuh takes the agent route. Its Syscollector module collects each endpoint's OS and package inventory, and the server correlates it with vulnerability data from the Wazuh CTI platform. A package is flagged when its version falls in a CVE's affected range. On Windows, enabling the hotfixes option lets Wazuh mark CVEs as fixed when an installed patch resolves them.
The advantage over a network scan: no scan account logging in to every host, and no scan window to schedule. The cost: assets without an agent are invisible. If you already run Wazuh as a SIEM in your lab, this is the cheapest scanner on the list to add; the open source SIEM guide covers the install.
OpenSCAP
OpenSCAP checks whether a host matches a hardening profile, rather than listing CVEs. It evaluates SCAP content such as the SCAP Security Guide profiles and writes an HTML report (getting started):
oscap xccdf eval --profile <profile-id> \
--results-arf arf.xml --report report.html \
/usr/share/xml/scap/ssg/content/ssg-<distro>-ds.xmlRun oscap info on the data stream file first to list its profile IDs. Reach for it when an auditor asks for evidence against a baseline.
What a vulnerability scan looks like from the SOC
Before you run any of these, you will probably triage one.
Network scans open connections from one source to many ports across many hosts in a short window. That is the pattern port scan rules for network service discovery (T1046) look for, and why scanner IPs end up on suppression lists.
Authenticated Windows scans log in to every host. Each successful login over SMB is a 4624 with logon type 3 (network), and the logon types guide explains why that type matters. When the scan account's password expires or gets locked, the same scan turns into a wave of event 4625 failures. Microsoft's 4625 reference says to watch sub status 0xC000006A (bad password) on service accounts, "especially" when several arrive in a row.
Note
Simulated example generated by SOCSimulator Research. An authenticated scan starts right after the scan account's password was rotated, but nobody updated it in the scanner.
| Time (UTC) | Computer | Event ID | Logon type | Account | Source IP | Sub status |
|---|---|---|---|---|---|---|
| 02:00:03 | FS-01 | 4625 | 3 | svc-vulnscan | 10.20.0.15 | 0xC000006A |
| 02:00:04 | WS-0412 | 4625 | 3 | svc-vulnscan | 10.20.0.15 | 0xC000006A |
| 02:00:04 | WS-0413 | 4625 | 3 | svc-vulnscan | 10.20.0.15 | 0xC000006A |
| 02:00:06 | SQL-02 | 4625 | 3 | svc-vulnscan | 10.20.0.15 | 0xC000006A |
| 02:00:07 | WS-0420 | 4625 | 3 | svc-vulnscan | 10.20.0.15 | 0xC000006A |
One account, one source, many hosts, seconds apart. In Microsoft Sentinel, this query surfaces the shape:
SecurityEvent
| where EventID == 4625 and LogonType == 3
| summarize Hosts = dcount(Computer), Attempts = count()
by TargetUserName, IpAddress, bin(TimeGenerated, 5m)
| where Hosts >= 10The query can't tell a broken scan from an attacker trying a stolen credential across the network. Your context can: is 10.20.0.15 in the scanner inventory, is 02:00 the scan window, and is svc-vulnscan the account the scanner uses? If any answer is no, escalate. The Windows event IDs cheat sheet lists the events around this one.
Web scans don't announce themselves. The User-Agent rules in older detection guides assume the scanner says its name. By default, none of the three web scanners here do:
| Scanner | Default User-Agent | Source |
|---|---|---|
| ZAP | A Chrome 141 on Windows string | ConnectionOptions.java in the network add-on |
| Nikto | The first entry of its db_useragents file, a Chrome 142 on Windows string; random entries with -evasion R | get_ua in nikto_core.plugin |
| Nuclei | A random browser User-Agent per request | build_request.go |
Nikto's nikto.conf.default already set a Chrome string in version 2.5.0; the old (Nikto/@VERSION) format survives only as a commented example. So match on source IP, request rate and the share of 404 responses, not on the User-Agent.
The best web signal is a callback. A DNS query for a subdomain of oast.pro, oast.fun or the other interactsh defaults, coming from one of your servers, means a Nuclei payload made that server reach out. If your team was scanning, that is a confirmed finding. If not, someone else is running Nuclei against you, and the template worked.
CI scanners talk to their databases. Trivy pulls from mirror.gcr.io and ghcr.io, Grype from grype.anchore.io, osv-scanner calls the OSV.dev API. With that list in hand, a runner calling scan.aquasecurtiy.org is a quick check instead of a guess.
Open source vs free: where Nessus Essentials fits
People searching for a free vulnerability scanner often end up on Nessus Essentials, so it belongs in the comparison even though it is not open source. Tenable's product page now describes it as a 30-day license for non-commercial use, capped at 5 IPs, with plugin updates delayed by 30 days. Nessus Essentials Plus raises that to 20 IPs with real-time plugins for $199 a year, and is free for verified students and instructors. Guides that still quote a 16-IP free tier are out of date.
Where to practise
Running a scanner in a home lab teaches you the tool. Triage is a separate skill: deciding in two minutes whether a 4625 burst is your scanner, a misconfiguration or an intrusion. The SOC tools guide shows where vulnerability scanners sit next to the SIEM, EDR and ticketing tools an analyst touches each shift.
Practice the triage side
Work realistic SIEM, XDR and Firewall alerts in a live SOC console. The free tier includes all three.
Frequently asked questions
Is OpenVAS free?
Yes. The scanner (openvas-scanner) is GPL-2.0, the manager (gvmd) is AGPL-3.0, and the Greenbone Community Feed of vulnerability tests is free. What you don't get for free is the Enterprise Feed, which Greenbone says adds tests for enterprise software and network gear such as Cisco, Microsoft Exchange and Palo Alto Networks, CIS Benchmark compliance checks and EPSS-based prioritization.
OpenVAS vs Nessus: which should I use?
For a lab or a learning project, OpenVAS, because it has no IP cap and the code is open. Nessus Essentials is free but proprietary: Tenable's page now describes it as a 30-day license for non-commercial use, limited to 5 IPs, with plugin updates delayed 30 days. Nessus Essentials Plus (20 IPs) costs $199 a year and is free for verified students and instructors. If your future employer runs Nessus, the Essentials Plus student license is the better way to learn its interface.
Is Nmap a vulnerability scanner?
Not on its own. Nmap finds hosts, open ports and service versions. With an NSE script such as vulners (nmap -sV --script vulners), it looks up the detected software versions in the vulners.com database and lists matching CVEs with CVSS scores. That is version matching, not testing, so treat the output as leads to verify, and use OpenVAS when you need actual vulnerability checks.
Is Trivy safe to use after the 2026 compromise?
Aqua's advisory (GHSA-69fq-xp46-6x23, CVE-2026-33634) lists what was affected: the Trivy v0.69.4 release, Docker Hub images 0.69.5 and 0.69.6, trivy-action tags before 0.35.0 and any setup-trivy tag (all 7, v0.2.0 to v0.2.6, were replaced; only SHA pins were safe). It says the malicious artifacts were removed, though copies may linger in caches. Current releases (v0.76.1 at the time of writing) are not on that list. Pin the action to a full commit SHA, verify binaries with cosign, and rotate any secret a pipeline exposed during the March 19 to 23 windows.
Is OWASP Dependency-Check still maintained?
Yes. The old jeremylong/DependencyCheck repository is archived because the project moved to the dependency-check organization on GitHub, where v13.0.0 shipped in August 2026 under Apache 2.0. Its README strongly recommends an NVD API key, since the tool switched from the NVD data feed to the NVD API in version 9.0.0.
Nerea writes SOCSimulator's guides for people starting or growing a career in security operations: how Windows authentication shows up in the logs (NTLM vs Kerberos), and which tools a T1, T2 or T3 analyst opens first. Each guide is built around the questions a new analyst actually has on shift, with real event IDs and queries to practice on.


