b4c2.ai

Same Kit, Four Misspellings

·b4c2 Research·2 min read
#ClickFix#Typosquatting#Threat Hunting

A one-hour hunt turned up four “verification” pages. Each pretends to be a different brand. All four are the same generated landing.

Four typosquatted brand verification lures generated from the same kit

Why This Still Works

Nobody types higgsflield[.]net on purpose. They get there from a 302 redirect, an ad, a “document shared with you” email, or a search result that looks close enough on a phone.

The page is a fake Cloudflare “Just a moment…” widget. It tells Windows users to press Win+R and paste; Mac users get Spotlight / Terminal instructions. The pasted line is not on the page in plaintext. It sits in window.__T[2] as base64. Decode it and every lure fetches the same host: leebin101[.]com.

So the attacker only has to:

  1. Register a lookalike domain.
  2. Stamp a new brand and token into a five-slot config array.
  3. Keep one dropper URL.

Rebuild cost is near zero. Domain cost is a few dollars. Human pattern-matching (“that looks like Claude / OpenAI / Cloudflare”) is still cheap to abuse.

What Did Not Change

Across all four scans, the raw kit strings were identical:

window.__T = [
  hideAllWidgetStates,
  (_0y = () => _0x && _0c),
  (_0x4926, 0xeade4),
];

Same obfuscator compile. Same C2 (leebin101[.]com).

Typosquatting is not a 2005 trick that “should be dead.” It is still the cheapest way to put a familiar name in the address bar for the three seconds it takes someone to paste a command they think is a captcha.

Scan Results

Considering this has potential to spread, here are the full un-obfuscated hunt results:

Un-obfuscated scan results across all four domains pointing to leebin101[.]com