Methodology
How we check and score
This page documents what a Webscan Radar scan does and how the score is calculated. The tables are generated from the same rule set the scanner uses, so page and scan cannot drift apart.
A scan looks at publicly visible configuration: one request for the start page, a few fixed paths, DNS records and connection attempts. It is a configuration check, not a security audit.
Scoring system
Every scan starts at 100 points. Each finding deducts the points listed in the tables below; the score is the starting value minus the sum of these deductions and never falls below 0. There are no hidden factors.
| Grade | Score |
|---|---|
| A | 90–100 |
| B | 75–89 |
| C | 60–74 |
| D | 40–59 |
| F | 0–39 |
Vulnerability or note
A finding with a deduction counts as a vulnerability, a finding without deduction as a note. Severity levels (Critical, High, Medium, Low, Info) indicate urgency and determine the order of the list; the points are fixed per rule and do not follow from the severity level.
When there is no grade ("n/a")
We assign no grade when a grade would be made up: (1) the site answers with a bot-protection challenge page instead of its content; (2) fewer than 50 % of the applicable core checks could run. The core checks are weighted: start page request 3, TLS connection 2, port check 1, e-mail records 1, subdomain lookup 1. Checks that do not apply (for example mail checks when an IP address is scanned) are not counted.
About the standards cited
The "basis" column names the standard that describes the checked property. The standards justify what we check. The number of points is our own weighting and is not taken from any standard. Where no standard fits, the column says "own assessment".
Checks and deductions
Transport encryption
Certificate. We open a TLS connection to port 443 and have the certificate chain and host name validated against the operating system's trust store. We read the issuer, expiry date, negotiated protocol version and cipher.
Remaining validity. Points are deducted only below 14 days of remaining validity. For issuers with customary automatic renewal (Let's Encrypt, ZeroSSL, Google Trust Services) it remains a note without points.
TLS 1.0 and 1.1. For each protocol we attempt a separate handshake offering only that version. Result: active, disabled or not testable (no answer) – we only report "disabled" when the server actively refuses.
Redirect from HTTP. We request http://<domain>/ without following redirects and assess the status code and redirect target. A closed port 80 counts as fine.
Mixed content. In the start page HTML we look for http:// addresses in script, iframe and link (stylesheet/preload) as active content, and in img, video, audio and source as passive content.
| Finding | Severity | Deduction | Basis |
|---|---|---|---|
| No valid HTTPS connection (invalid certificate or connection fails) | Critical | −30 | Mozilla Web Security Guidelines (HTTPS); BSI TR-02102-2 |
| Certificate expired | Critical | −30 | RFC 5280 |
| Certificate expires in less than 14 days | High | −10 | own assessment |
| Certificate expires in less than 14 days, issuer with customary automatic renewal | Info | 0 (note) | own assessment |
| IP address scanned: no certificate to check | Info | 0 (note) | own assessment |
| TLS 1.0 is accepted | High | −10 | RFC 8996; BSI TR-02102-2 |
| TLS 1.1 is accepted | High | −10 | RFC 8996; BSI TR-02102-2 |
| Site is fully reachable over unencrypted HTTP (status 200) | Critical | −15 | Mozilla Web Security Guidelines (HTTP Redirections) |
| HTTP redirects, but not to HTTPS | High | −10 | Mozilla Web Security Guidelines (HTTP Redirections) |
| Active content (scripts, frames, stylesheets) loaded over HTTP on an HTTPS page | Critical | −15 | W3C Mixed Content |
| Images/media loaded over HTTP on an HTTPS page | Medium | −5 | W3C Mixed Content |
HTTP headers and cookies
Security headers. We fetch the start page (at most 2 MB, up to 5 redirects) and assess the headers of the final response. HSTS counts as set when max-age is greater than 0; X-Frame-Options counts as satisfied when the CSP directive frame-ancestors is set instead.
Content-Security-Policy. If a CSP is present we check four core directives (object-src may be covered by default-src) and whether scripts allow 'unsafe-inline' without a nonce, hash or 'strict-dynamic'. We do not evaluate other policy weaknesses.
Cookies. We read the Set-Cookie headers of the start page response and check each cookie for the Secure, HttpOnly and SameSite attributes. We do not see cookies that are only set by JavaScript.
security.txt. We request /.well-known/security.txt and show whether the file exists with a Contact entry (RFC 9116). The result does not affect the score.
| Finding | Severity | Deduction | Basis |
|---|---|---|---|
| Strict-Transport-Security missing or max-age=0 | Medium | −10 | RFC 6797; OWASP Secure Headers Project |
| Neither X-Frame-Options nor CSP frame-ancestors set | Medium | −10 | RFC 7034; W3C CSP Level 3 (frame-ancestors) |
| X-Content-Type-Options missing | Medium | −10 | OWASP Secure Headers Project; Mozilla Web Security Guidelines |
| Content-Security-Policy not set | Medium | −5 | W3C CSP Level 3; OWASP Secure Headers Project |
| CSP present but core directives missing (object-src, base-uri, frame-ancestors, form-action) | Low | −2 | W3C CSP Level 3 |
| CSP allows 'unsafe-inline' for scripts without nonce/hash | Info | 0 (note) | W3C CSP Level 3 |
| At least one cookie without Secure or HttpOnly flag | High | −10 | RFC 6265; OWASP Session Management Cheat Sheet |
| Cookies only lacking the SameSite attribute | Low | −3 | RFC 6265bis (Entwurf) |
E-mail sender protection
Mail operation. We first check the MX records. If the domain has no MX or a null MX (RFC 7505) it does not send e-mail – missing SPF/DMARC records then cause no deductions.
SPF. We read the domain's TXT records and assess the all mechanism of the SPF record: -all and ~all count as effective, ?all or no all as ineffective, +all as open. We do not follow redirect= and show that as "not assessed".
DMARC. We read the TXT record at _dmarc.<domain> and assess the policy: quarantine and reject count as enforced, none as monitoring only. We do not evaluate the subdomain policy (sp), pct or alignment.
| Finding | Severity | Deduction | Basis |
|---|---|---|---|
| No SPF record (domain handles mail) | High | −15 | RFC 7208 |
| Multiple SPF records (invalid) | High | −15 | RFC 7208, Abschnitt 4.5 |
| SPF ends in +all | High | −15 | RFC 7208 |
| SPF ends in ?all or has no all mechanism | Medium | −8 | RFC 7208 |
| No DMARC record (domain handles mail) | Medium | −10 | RFC 7489 |
| DMARC record invalid (multiple records or no valid policy) | Medium | −10 | RFC 7489 |
| DMARC policy p=none | Low | −5 | RFC 7489 |
| Domain without mail and without SPF: recommendation "v=spf1 -all" and null MX | Low | 0 (note) | RFC 7208; RFC 7505 |
Exposed services
Port check. We attempt a TCP connection to 20 ports of the domain's first IPv4 address (21 FTP, 22 SSH, 23 Telnet, 25 SMTP, 53 DNS, 80 HTTP, 110 POP3, 143 IMAP, 443 HTTPS, 445 SMB, 1433 MSSQL, 1521 Oracle DB, 3306 MySQL, 3389 RDP, 5432 PostgreSQL, 5900 VNC, 6379 Redis, 8080 HTTP-Alt, 8443 HTTPS-Alt, 27017 MongoDB). We treat 23 (Telnet), 445 (SMB), 1433 (MSSQL), 1521 (Oracle DB), 3306 (MySQL), 3389 (RDP), 5432 (PostgreSQL), 5900 (VNC), 6379 (Redis), 27017 (MongoDB) as critical. If the address belongs to a detected CDN we say so and deduct no points.
FTP. If port 21 is open we read the server greeting and send a user name – never a password. If the server accepts that without TLS, unencrypted login is possible; if it demands TLS first, we rate it lower.
| Finding | Severity | Deduction | Basis |
|---|---|---|---|
| Database, remote administration or legacy protocol port reachable from the internet (once, regardless of count; no points behind a CDN) | Critical | −20 | own assessment |
| FTP: login without TLS possible | High | −10 | RFC 4217 |
| FTP open, login only with TLS | Medium | −5 | own assessment |
| FTP open, login method could not be checked | Medium | −5 | own assessment |
Publicly retrievable files and information
Files and directories. We request 13 typical paths (/.env, /phpinfo.php, /info.php, /test.php, /backup.sql, /backup.zip, /database.sql, /wp-config.php, /.DS_Store, /.htaccess, /server-status, /server-info, /config.php) as well as /.git/HEAD. We only report a finding when the content shows the leak (for example a valid Git reference, PHP source code, an Apache status page); a normal HTML page at the path does not count.
robots.txt. From robots.txt we request up to 12 disallowed paths with conspicuous names (such as admin, backup, internal). Only an open directory listing or a directly retrievable dump is critical; protected standard CMS paths are not reported.
Server banner. We evaluate the Server and X-Powered-By headers. A finding only arises for a version number (digit.digit) or when X-Powered-By is set.
E-mail addresses. We search the start page source for e-mail addresses. Placeholders and example addresses (such as in placeholder attributes or at example.com) do not count.
Subdomains. We query 20 common names (www, mail, admin, staging …) via DNS and detect wildcard records. The result is only displayed and does not affect the score.
| Finding | Severity | Deduction | Basis |
|---|---|---|---|
| .git directory retrievable | Critical | −25 | OWASP WSTG-CONF-04 |
| Sensitive file retrievable (content verified) | Critical | −10 each, max. −30 | OWASP WSTG-CONF-04 |
| Standard path returns content without a clear marker | Low | 0 (note) | OWASP WSTG-CONF-04 |
| Path listed in robots.txt returns an open directory or dump | Critical | −8 each, max. −20 | OWASP WSTG-INFO-03 |
| robots.txt lists non-standard paths | Info | 0 (note) | OWASP WSTG-INFO-03 |
| Server or X-Powered-By header reveals a version number | Low | −2 | OWASP WSTG-INFO-02 |
| E-mail address in the start page source | Low | −3 | own assessment |
Software currency
CMS detection and version. We detect common CMS and shop systems by markers in the HTML and read the version from the generator meta tag (for WordPress alternatively from version parameters of specific files). We compare against a daily updated list of current versions; without a detectable version there is no finding.
Known vulnerabilities (CVE). We only match against a small, individually verified list: CVE-2024-21726 (Joomla), CVE-2024-34356 (TYPO3), CVE-2023-22731 (Shopware). A finding arises only when the version was fully detected and lies within the affected range of the CVE record. We do not check plugins or extensions.
| Finding | Severity | Deduction | Basis |
|---|---|---|---|
| Detected CMS version is at least one major version behind | High | −15 | OWASP Top 10 2021 – A06 |
| Detected CMS version is more than two minor versions behind | Medium | −8 | OWASP Top 10 2021 – A06 |
| Verified CVE (critical) for the detected version | Critical | −20 | CVE-Record (cve.org); OWASP Top 10 2021 – A06 |
| Verified CVE (high) for the detected version | High | −10 | CVE-Record (cve.org); OWASP Top 10 2021 – A06 |
| Verified CVE (medium) for the detected version | Medium | −5 | CVE-Record (cve.org); OWASP Top 10 2021 – A06 |
Operations
Domain expiry. We query the domain's expiry date via RDAP at the registry. For endings without open RDAP access (including many country-code domains) the check is skipped.
Response time. When fetching the start page we measure the time until the response headers arrive. This is a single measurement from our server, not a load test.
| Finding | Severity | Deduction | Basis |
|---|---|---|---|
| Domain registration expired | Critical | −30 | own assessment |
| Domain expires in less than 30 days | Critical | −25 | own assessment |
| Domain expires in less than 60 days | Medium | −5 | own assessment |
| Time to first byte 2500 ms or more | Medium | −8 | own assessment |
| Time to first byte 1500 ms or more | Medium | −5 | own assessment |
Privacy notes
What we look at. We search the start page source (for pure JavaScript applications additionally up to three first-party script files) for markers of known tracking services, known consent tools, Google Fonts embeds and the words imprint and privacy.
What follows from it. The privacy indicator (0–100) is separate from the security score and uses fixed weights: tracking without a detected consent tool −25, word "privacy" not found −20, word "imprint" not found −15, Google Fonts loaded externally −15. We cannot tell whether a tracker only loads after consent – we describe observations and make no legal assessment.
| Finding | Severity | Deduction | Basis |
|---|---|---|---|
| Tracking embed found, no known consent tool detected | Low | 0 (note) | own assessment |
| Google Fonts loaded from Google servers | Low | 0 (note) | own assessment |
What we do not check
- No JavaScript rendering: we read the delivered HTML. Content, trackers and cookies that only arise in the browser are invisible to us.
- Only the start page and a few fixed paths – no subpages, no crawling.
- No in-depth TLS analysis: no cipher suite enumeration, no check of key size, signature algorithm, OCSP stapling or Certificate Transparency. For that we recommend Qualys SSL Labs.
- No DKIM in the full scan (the single-purpose tool tries nine common selectors), no DNSSEC, MTA-STS, TLS-RPT or DANE, no assessment of CAA records.
- For HSTS we assess neither the duration nor includeSubDomains or preload; Referrer-Policy and Permissions-Policy do not affect the score.
- No active attack tests (such as for XSS or SQL injection), no malware detection, no blocklist lookups.
- No detection of plugins, themes or libraries and their versions.
- Port check via TCP only, IPv4 only, only the first address of the domain; no UDP, no service detection except for FTP.
- Behind a CDN we test its edge servers, not the origin server. If bot protection serves a challenge page, we assign no grade.
- No legal review: the privacy notes are no substitute for a legal assessment, and a good grade does not certify conformity with any law or standard.
- Every result is a snapshot at the time of the scan and no substitute for a security audit.
Data handling
- Per scan we store the address checked, the result, the time and a screenshot of the start page of the checked website; for signed-in users the link to the account.
- We store the IP address from which a scan was started to prevent abuse and automatically remove it from the scan after 7 days. We determine the country of origin exclusively locally on our server ("IP to Country Lite" database by DB-IP, CC BY 4.0); the IP address is not transmitted to anyone for this. The country remains stored as a statistical value.
- Scans without an account are deleted automatically 12 months after the scan, including the screenshot. Scans of signed-in users remain until the user deletes the scan or the account.
- We keep daily database backups for 60 days; data that has already been deleted may still be contained in them until that period expires.
- With an account up to 3 domains whose ownership has been verified by a DNS record can be monitored: we scan them daily and notify by e-mail when a new finding of severity Critical or High appears or the score drops by 10 points or more. These scans run without a screenshot. We keep the monitoring scans of the last 90 days plus every scan that triggered a notification.
- Results of the single-purpose tools are cached for 24 hours. When a report is requested as a PDF by e-mail, we store the request for 365 days.
- A result is accessible via its unguessable link. Visitors without an account see the title and severity of all findings but details of only one; discovered paths, addresses and versions are not delivered to them.
- During a scan we query third parties: DNS records via Cloudflare (DNS over HTTPS) and the domain's expiry date at the responsible registry (RDAP). Only the checked domain is transmitted. The icon of the checked website is loaded from Google by the viewer's browser.
- Per scan the checked website receives: one start page request, up to three script files for JavaScript applications, 17 fixed paths, up to 12 paths from robots.txt, connection attempts to 20 ports, three TLS connections, one request over HTTP and one page load for the screenshot.
The scanner identifies itself. Its requests carry this User-Agent; the browser that takes the screenshot appends the same name to its own identifier:
WebscanRadar/1.0 (+https://webscan-radar.com/methodik)
If you operate a website and do not want it to be checked, you can block this identifier on your server or write to us at office [at] webscan-radar.com.
Legal details are in the privacy policy.
Changelog of the rules
2026-11-14
- CVE matching reduced to entries individually verified against cve.org; five misattributed identifiers removed. Without a fully detected version there is no CVE finding any more.
- TLS 1.0/1.1: the check now actually measures; when it cannot, the result is "not testable" instead of "disabled".
- SPF and DMARC are assessed by content (all mechanism, policy) rather than mere presence; full scan and single-purpose tool share the same assessment.
- File findings only after content verification; placeholder addresses no longer count as e-mail findings; paths are checked against the host that actually serves the page.
- X-Frame-Options counts as satisfied when the CSP directive frame-ancestors is set.
- The score is exactly the starting value minus the sum of the listed deductions. Privacy notes no longer carry points.
- Privacy findings are worded as notes, without legal judgement.
- FTP is no longer critical across the board: high only when login without TLS is possible, otherwise medium.
- Certificate validity: deduction only below 14 days (previously below 60 and 30 days); for issuers with automatic renewal only a note.
- Findings without a deduction are counted as notes, not as vulnerabilities.
- The scanner identifies itself with its own user agent (previously a browser string). The country for an IP address is now determined locally only; the IP address is removed from the scan after 7 days, scans without an account are deleted after 12 months.
- Stored older results were converted to these rules; grades may have changed as a result.