← All articles

WordPress security: one scanner drove a .tw attack surge

WordPress security: one scanner drove a .tw attack surge

WordPress security data from the Wordfence Threat Intelligence team links a sharp rise in attacks on Taiwanese websites to one IP address, 45.38.60.254. An IP address is the numeric network address used to identify a device or service. The increase began in late October 2026, when the number of .tw websites attacked each day approximately tripled. Wordfence found no comparable change across .jp, .kr, .hk, .sg or .cn websites.

Wordfence’s records show that 45.38.60.254 began scanning on October 26th, 2026. The Wordfence Firewall, a system that examines web requests and blocks known malicious patterns, has blocked or logged more than 246,000 requests from that address. Of the requests stopped by the Malicious File Upload rule, 82.7% targeted websites using the .tw country-code top-level domain, which is the ending used by Taiwanese websites. According to Wordfence’s logs, the requests attempted to exploit known vulnerabilities and were stopped by generic firewall rules.

To measure the regional change, Wordfence counted websites where its firewall blocked at least one request per day between September 1st and November 15th, 2026. The review covered .tw for Taiwan, .jp for Japan, .kr for South Korea, .hk for Hong Kong, .sg for Singapore and .cn for China. It examined five general rules: SQL Injection, which targets weaknesses in database queries; Cross-Site Scripting, which attempts to place harmful code into pages; Malicious File Upload; Malicious File Upload (Patterns); and Malicious File Upload (PHP). All other domains combined served as a global baseline.

Counting affected websites rather than requests was deliberate. One attacker can send a very large number of requests to one site without showing that activity has spread across a region. The number of affected websites changes only when the activity reaches more sites. Five of the six country-code domains showed no significant change during the period. The exception was .tw.

From October 26th through November 13th, an average of 426 .tw websites per day had requests blocked by the Malicious File Upload rule, compared with an average of 131 during the previous eight weeks. SQL Injection showed the same pattern: the daily average rose from 216 .tw websites to 599 during the same period. The figure then fell to 102 websites on November 14th and 97 on November 15th.

Compared with September 1st through October 25th, the average number of attacked .tw websites during October 26th through November 13th was 3.25 times higher for Malicious File Upload and 2.77 times higher for SQL Injection. For the other country-code domains, those ratios remained between 0.69 and 1.35. Malicious File Upload (PHP) also increased on several domains, but the sevenfold rise on .tw was by far the largest.

The IP address at the centre of the activity triggered the Malicious File Upload (PHP) rule 831 times across 403 .tw websites and 238 times across 122 websites under other top-level domains. It did not trigger that rule on any .jp, .kr, .hk, .sg or .cn website. The next two addresses in the ranking sent 7,595 and 2,322 comparable requests to the rest of the world, despite hitting .tw 328 and 113 times respectively. Wordfence therefore describes those addresses as worldwide scanners, while 45.38.60.254 almost exclusively targeted .tw sites.

On the three days when the rule reached its highest levels on .tw, 45.38.60.254 reached most of the affected websites. From November 1st through 13th, the daily counts of attacked .tw websites added up to 5,541, of which the address accounted for 3,732, or 67%.

The address reached between 568 and 863 Wordfence-protected websites every day from October 26th through November 13th. From November 14th onward, that number dropped to 15 or fewer. In November, Wordfence blocked requests from the address under these generic rules: SQL Injection, 29,399 requests across 1,093 websites; Directory Traversal, 21,870 across 1,022; Malicious File Upload, 13,563 across 771; Local File Inclusion, 5,812 across 680; Cross-Site Scripting, 4,691 across 664; Directory Traversal via HTTP Headers, 2,957 across 579; and Malicious File Upload (PHP), 2,957 across 525.

Wordfence examined the two Malicious File Upload categories in detail. Its logs indicate that all the requests attempted to exploit previously disclosed vulnerabilities in WordPress plugins and other web software. The CVE identifiers in the request payloads ranged from 2017 to 2024. CVE is the standard identification system for publicly disclosed software vulnerabilities. The uploaded files did not contain a web shell or backdoor. Instead, they contained a short PHP snippet that printed a hash, allowing the scanner to check whether an upload had succeeded.

One example targeted `/jquery-file-upload/server/php/index.php` and attempted to upload a file named `dQFMK0WG.php`. The PHP code printed an MD5 hash associated with CVE-2018-9206. Wordfence says the request was blocked, and that no vulnerability-specific firewall rule was needed because its generic rules stopped the activity.

Of the 771 websites where Malicious File Upload requests from 45.38.60.254 were blocked, 610, or 79%, used .tw domains. Another 129, or 17%, used .com domains. Based on their domain names, Wordfence says the targeted .com sites also appeared to have ties to Asia. The evidence suggests the activity was primarily mapping Taiwanese websites, including WordPress sites, for known vulnerabilities rather than representing a broad increase in attacks across every Asian domain.

For site owners, the incident shows why old plugin and other web-software vulnerabilities remain relevant even when an attack sends only a small test file. Keep WordPress plugins and other installed software updated, review unusual upload activity, and use a firewall that can block generic malicious requests. Maintain current backups so a confirmed compromise can be investigated and the site restored. A managed service such as AEU Hosting provides managed WordPress hosting secured end to end, which is directly relevant to reducing the exposure of a WordPress site to this kind of probing.

Wordfence’s review found no evidence that the inspected uploads contained a web shell or backdoor, but it describes the activity as reconnaissance, meaning an attempt to map sites and identify vulnerable targets. The rise in .tw attacks was therefore largely the result of one scanner covering one list of sites, not evidence of a region-wide increase in hostile activity. Wordfence said it would continue watching for the same scan from other addresses.

Also available in: DeutschItalianoShqip