· ~5 min read
What hit my static site in its first week online
Fifteen thousand requests in five days, more than half of them looking for files that do not exist: .env files, git configs, WordPress installers and cloud credentials. A breakdown of who came, what they tried, and why none of it worked.
Why look
This site moved to its own domain, viotnet.com, on 7 November. A day later a "new website" directory emailed me to say it had found it. That made me curious about who else had. Cloudflare sits in front of the site, so I pulled the last week of edge analytics with a read-only API token and sorted through it.
The window is 7 to 12 November 2026, a little over five days. The numbers come from Cloudflare's analytics for the zone; on the free plan, detailed per-request data is sampled and kept only for a short time, so treat them as close estimates, not exact counts. I am not publishing any IP addresses: they are personal data, and the countries and tools tell the story well enough.
The headline numbers
| Requests | Share | |
|---|---|---|
| Total | 15,470 | 100% |
| 404 Not Found | 8,163 | 53% |
| 301 redirects (www and the old pages.dev address) | 4,038 | 26% |
| 200 OK | 2,793 | 18% |
| Blocked or refused (403, 405) | 370 | 2% |
| Timeouts on ports the site does not serve (5xx) | about 470 | 3% |
More than half of everything that reached the site asked for a page that has never existed. The real readers are in the 200s: the home page about 1,900 times, each post between 35 and 45 times, and the Greek pages a few dozen times each.
Who came
Sorted by the User-Agent each client sent, which a client can fake, so this is what they claimed to be:
| Kind | Requests | Examples |
|---|---|---|
| Looks like a browser | 8,385 | many are scanners using a browser string |
| Declared security scanners | 5,004 | "TLM-Audit-Scanner" (4,483), LeakIX (369), Censys, internet-measurement.com |
| Scripts and libraries | 831 | curl (594), Go, python-requests, aiohttp, okhttp |
| AI crawlers | 146 | GPTBot (66), ClaudeBot (63) |
| Search engines | 138 | Googlebot |
| SEO crawlers and link previews | 41 | serpstatbot, Facebook |
One client did not even bother with a believable User-Agent: it sent the URL of a WordPress installer as its name, 490 times.
By country, the Netherlands led with about 6,400 requests, then the United States, Singapore, Germany and Japan. That does not mean Dutch people love this blog: those countries host a lot of cheap cloud servers, which is where scanners run from. The failed requests (4xx) came mostly from the Netherlands, Singapore, Russia and the US.
What they were looking for
Grouping the 4,700 failed requests by what the path was after:
| Target | Requests | Typical paths |
|---|---|---|
| Secrets and config files | 2,252 | /.env, /.env.production, /.env.local, /.git/config, /.aws/credentials, /config.json |
| WordPress | 987 | /wp-admin/install.php, /wp-login.php, /xmlrpc.php, wlwmanifest.xml under 15 different folders |
| PHP files and shells | 275 | /info.php, /phpinfo.php |
| Admin panels | 137 | login and management pages |
| Application internals | 79 | /actuator/env (Spring Boot) |
A few patterns stand out:
- Leaked secrets are the most wanted thing on the internet. Over a third of all failed requests looked for environment files, git metadata or cloud credentials. A
.envfile accidentally deployed with a site is a common source of leaked API keys and database passwords, and bots check every new site for one within days. They tried.envwith every suffix a tired developer might use:.save,.backup,.prod,.dev, and insideapi/,src/andcms/. - WordPress installer hijacking.
/wp-admin/install.phpwas the single most requested missing page. If someone uploads WordPress and does not finish the setup, whoever reaches the installer first becomes the admin. Bots race to be first. - WordPress discovery by brute force. The same file,
wlwmanifest.xml, was requested under/wp/,/blog/,/shop/,/2019/,/test/and ten more folders, about 35 times each: a scanner guessing where a WordPress install might live. - Every port, not just 443. Requests also arrived on 8443, 8080, 2053, 2083, 2087 and others, the alternative ports Cloudflare can proxy. Nothing listens there for this site, so they ended in timeouts. Those are the 5xx errors in the table above, and they are harmless.
What Cloudflare stopped
Cloudflare's free managed rules blocked about 260 requests outright before they reached the site. Most were looking for backups of a WordPress config file: wp-config.php.bak, .old, .save, .swp, ~, .orig. These are the copies an editor or a hurried admin leaves behind, and they contain the database password in plain text.
Why none of it worked
This site is static: plain HTML and CSS files, no server-side code, no database, no admin panel, no JavaScript and no forms. There is no .env to leak because nothing on the server reads environment variables. There is no WordPress to hijack. A POST request has nothing to talk to and gets a 405. Every probe above ended in a 404, a redirect to a 404, or a block. I checked the most-requested paths by hand afterwards to make sure.
That is the real lesson. The most effective defence here was not a firewall rule, it was having almost nothing to attack. The attack surface of a static site is the hosting platform and the account that controls it, which is why the Cloudflare account sits behind a hardened Google login with two-factor authentication.
What I am changing
- Nothing urgent. No finding needs a fix.
- Less noise. I may add a custom firewall rule that blocks requests for
.env,.git,wp-paths and.phpfiles at the edge, so they never show up as 404s. It changes nothing about security, but it makes the real traffic easier to see. - A favicon.ico. Browsers ask for
/favicon.icoby default, and it was one of the more common 404s from real visitors. - AI crawlers. GPTBot and ClaudeBot both read the site. I have not decided whether to allow that in
robots.txt; for a public blog, I lean towards yes.
If you run anything on the public internet, look at your own logs for .env and .git requests. The bots already have.
Want the configs or code from this lab, or have a question? Email me at papas@viotnet.com