PILOTPoland pilot. We publish aggregate findings, never the names of compromised sites.
HackedIndex Poland · pilot Check your site

Methodology · Poland pilot

What we know about a hack.
And what we do not infer.

HackedIndex looks for evidence of spam on publicly accessible websites. This page explains what enters the index, how we calculate the figures and where an external scan reaches its limits.

Scope

A defined view of the Polish web.

The pilot covers a domain list connected to the Polish market. We use public HTTP Archive and Chrome UX Report datasets alongside store lists. A Polish domain suffix or observed traffic from Poland can qualify a domain. Neither establishes that its owner or hosting provider is Polish.

This list is neither a census of Polish websites nor a random sample of the whole internet. Each source has its own visibility thresholds and tends to include popular websites and recognisable platforms. Our rates describe the scanned list, rather than the prevalence of hacking across every website in the country.

The vulnerability database focuses on WordPress and PrestaShop.

We match vulnerabilities only to publicly identified versions. When a component's version is not visible, its presence alone does not establish vulnerability. No match does not mean the component is safe.

Public content checks

We request public pages using a browser user agent and a Googlebot user agent. We inspect response codes, redirects, HTML and sitemaps. A scan may also visit selected internal pages.

No access to the back office

Requests identify the scanner as HackedIndex. It does not sign in, submit forms or attempt to exploit vulnerabilities. It has no access to non-public server files, databases or administrator logs.

Evidence before suspicion

Confirmed evidence of compromise enters the count.

We look for unwanted content that uses someone else's website to promote casinos, betting, pharmaceuticals or other spam. A casino keyword, an old plugin or a suspicious URL alone is not enough.

Hidden spam and known patterns

Strong evidence includes a hidden spam block, a recognised spam network pattern or a clear replacement page served to a bot. When comparing responses, we account for error pages and anti-bot challenges.

Independent corroboration

A spam link needs an additional signal, such as sitemap entries, spam in Google results or a known spam host. Content shown only to a bot requires corroboration in Google.

Sitemaps in context

An isolated suspicious URL is not proof. We consider the number of entries, their paths and their context. Rules can recognise a mass injection of URLs; weaker cases need additional confirmation in Google results.

Unclear findings stay separate

Signals without sufficient confirmation do not increase the compromised-site count. We distinguish them from no evidence found, access blocked and a site that did not respond.

Vulnerability, compromise, cause

Different findings need different evidence.

A known vulnerability

A visible software version falls within an affected range described by a security source. It calls for an update or a configuration check. It does not establish that the flaw was exploited on that website.

A detected compromise

A public response contains confirmed evidence of injected spam. We can detect compromise without knowing plugin versions or the entry point.

An established cause

Attributing a compromise to a specific vulnerability needs additional evidence, such as logs, file analysis or a substantiated incident report. A plugin's presence, including a Pro edition, does not establish that link.

We read versions from metadata, HTML comments, asset URLs and public software files. We discard recognised placeholder values, timestamps and library versions that could be mistaken for a plugin version. If a version is not visible, it remains unknown. We do not estimate it from the appearance of the site.

Matching considers the product, affected version ranges and available patch information. Hidden components, modified files, backported fixes and configuration-dependent conditions limit the certainty of the assessment. A version check does not attempt exploitation.

Explore the vulnerability radar

Counting rules

A figure needs its date and denominator.

  • We count domains. A domain enters a scan once. Many infected pages do not become many compromised domains.
  • The rate covers sites that returned content. Its denominator includes confirmed spam, unclear findings and no evidence found. Unreachable sites and sites blocking the scan are excluded; exclusion does not mean they are clean.
  • Techniques overlap. A website can hide links, replace content for Googlebot and contain sitemap spam at the same time. Technique totals therefore do not add up to the domain count.
  • The scan date is an observation date. It does not tell us when an intrusion happened. Similarly, an absence of spam in a later scan does not prove a backdoor was removed.

A changed list or method changes the comparison.

The September study covered domains with Polish traffic. A later scan expanded the list and began calculating the rate against sites that actually returned content. We mark this change on the chart and show a calculation using the earlier measure where available.

Confirmation coverage also differs: the September study used Google results, while a weekly scan can run without that check. Findings that need this source then remain unclear. A lower rate after a methodology change does not establish a decrease in hacking.

Different datasets have different update schedules.

The domain scan is scheduled weekly, while the known-vulnerability database refreshes daily. A newly published flaw can be matched to a previously observed version without revisiting a website. The database update date is therefore not a new scan date for every domain. Figures identify the date of their underlying data.

Limitations

No evidence found is not a security certificate.

We inspect a server's response at a particular time and from a particular address. We do not visit every page or execute the site as a full browser would. Spam loaded only by JavaScript, or triggered only for a particular country, device, referrer or IP address, may remain invisible.

A Googlebot user agent does not make the scanner a real Google crawler. A website can also check a visitor's network and address. Access controls, server errors, time limits, redirects and caches may restrict what we see. Search results can retain spam that the owner has already removed.

A private report gives an administrator or hosting provider a starting point. Establishing that the root cause has been removed requires back-office access and incident analysis.

Check your website

Privacy and responsible disclosure

Public figures. Private details.

We do not publish the names of compromised domains, including public institutions. The public site shows aggregated findings and descriptions of patterns. Detailed website reports require verification of control over the domain.

Evidence intended for disclosure goes privately to the owner, the relevant response team, CERT Polska in this pilot, and the hosting provider. Detection does not establish that a notice has already been sent or read. Those are separate facts that need confirmation.

HackedIndex is a side project by Mateusz Haber. It observes and alerts; it does not sell website cleanup or collect referral commissions.

Read about our notices

Sources and interpretation

A security source is not an incident report.

Vulnerability records describe products and versions. CISA KEV establishes known exploitation in the wild, not exploitation of every website running that software. EPSS estimates exploitation probability; it does not confirm compromise.