20 Nov 2026 / Darknet News
Law Enforcement Roundup 2026: Directory Spoofing Arrest Patterns
The most interesting arrest pattern I’ve tracked over the past eighteen months has almost nothing to do with darknet markets themselves. It shows up in the infrastructure that markets, carding shops, and phishing crews all rent from the same handful of providers: domains, redirects, anti-detect profiles, and hosted phishing kits. Directory spoofing, the practice of dressing up a malicious link as a legitimate platform URL, has become the connective tissue in a surprising number of 2025 and 2026 prosecutions. The arrests cluster around it, and the technical profile of the people getting caught is different from the profile of the people running the markets.
What directory spoofing actually means in the case files
Strip the jargon and it’s simple. An attacker builds a redirect chain that looks like normal business traffic. A PDF lands in an inbox titled something like “Final Lien Waiver.pdf.” The victim clicks a QR code inside it. Traffic passes through an open redirect on a legitimate domain, then hits a Cloudflare Turnstile gate disguised as a benign page, in one documented case a fake food site called “Gourmet Delights.”
That is directory spoofing at the infrastructure layer. The URL the victim sees belongs to a trusted service. The payload lives somewhere else. And crucially, IP-based filtering quietly shoves data-center, VPN, proxy, and known-abuse IPs off to a Microsoft Wikipedia page via href.li. Researchers get a decoy. Real victims get the fake login.
That last detail matters more for law enforcement than it first appears. The spoofing layer is a form of defensive camouflage against automated detection systems. The same filtering logic that evades a security scanner also evades the crawlers and sandboxes that threat intelligence teams rely on. When those sandboxes get decoyed, investigators lose the easy early warning. That slows attribution and pushes enforcement toward the harder work of infrastructure seizure and human intelligence.
The phishing kit layer where arrests actually land
Three PhaaS platforms dominate the current case volume: Sneaky 2FA, EvilTokens, and EvilProxy. Each uses a different technical mechanism. All three converge on the same outcome, a fully authenticated M365 session or token in the attacker’s hands without ever cracking MFA. Sneaky 2FA and EvilProxy sit as a live man-in-the-middle proxy, relaying the real authentication flow and skimming the session cookie the moment Microsoft issues it.
The operational detail that makes these platforms enforcement-friendly is that they leak. Sekoia’s full IOC set for Sneaky 2FA spans roughly 61 indicators: 57 domains, two IPs, and two subdomains, published in CSV form through the SEKOIA-IO Community GitHub repository. Sixty-one indicators is a lot of provider relationships for a crew to maintain. Every domain is a registration record. Every IP is a billing trail.
EvilTokens adds a second kind of footprint. The device-code flow it abuses requires an attacker to first check whether the target account even exists, using the GetCredentialType endpoint on login.microsoftonline.com. Push Security and Sekoia both documented multiple shared backend IPs on the Railway cloud platform used for token replay. Shared hosting is the single most common thread in the 2026 arrest wave. Six different crews using the same Railway IP block is a gift to any investigator with subpoena power.
The “impossible device shift” as a forensic artifact
There is one detection signal I think will end up in more charging documents than it will in corporate AV logs. Sneaky 2FA hardcodes a different User-Agent string for each step of the authentication flow it relays. A Safari-on-iOS User-Agent for the login step, then a Chrome-on-Windows User-Agent for MFA resolution. No real browser produces that sequence inside one session.
Sekoia calls it the impossible device shift, and it is detectable via Sigma correlation rules against Entra ID/M365 audit logs, correlating Login:login and Login:resume events by correlation ID within a ten-minute window. For a prosecutor, that is a timestamped, provider-held log of a specific attack session. It is not anonymous. It ties a session to an account to a location to a billing identity, in that order.
The anti-detect fingerprint problem
Directory spoofing in the carding context leans on fingerprint matching. If your spoofed user agent says Windows 10 but your font list includes macOS-specific fonts like Helvetica Neue or Menlo, that is a mismatch. A card from Miami with a browser timezone set to New York is flagged. Language headers must be exact: en-US for a US card, not en-GB, not fr-FR. WebRTC must be disabled entirely or forced through the proxy, because even a SOCKS5 proxy can leak the real IP through STUN requests.
This is the part where the arrest pattern gets ugly for defenders. The fingerprint layers take hours to align manually. Most operators don’t. They buy pre-configured profiles that ship with canvas hash, WebGL fingerprint, audio context, fonts, timezone, language, and screen resolution all matched to the cardholder’s geo before a single tab opens.
That convenience is also a single point of failure. Pre-configured profiles are a centralized service. If you can identify the profile provider, you have identified a large fraction of the operators using that profile. The 2026 arrest clusters reflect exactly this structure: law enforcement did not find every carding crew individually. They found the profile vendor and worked backward through the subscriber list.
| Torzon Market |
torzon7aphar3x4l5b77nsylgyw26kntbi4m2wemrjh72aczeh27f6qd.onion
|
| DrugHub |
drughuj7l72ig56pza77eriu7yh6qsao4xb4yasq2qfjusxzuq6rlwqd.onion
|
| DarkMatter |
darkmafmzgnsmow5z3spgludhpwxhwbg77oam433fjx3clzh2yp2oaid.onion
|
| Omega Market |
omega7yhz7n4vg4yhf2na2qaaaeatdlqvjbj2juc245mr5muxtnuvgyd.onion
|
| BlackOps |
blackoogcnxogvymmebfwfjhx4k7efpgeoeytxtsev2lc4pqlbz54qad.onion
|
Why the encrypted-comms platform arrests keep occurring
The other structural weakness is the assumption that encrypted communication platforms are safe. Authorities have increasingly penetrated supposed-secure criminal communications platforms, and the pattern is consistent. Investigators analyze the seized data, arrests follow, and additional arrests come in the next months as the data is worked through. The March 2025 takedown of an encrypted comms platform, which led to arrests across multiple jurisdictions and continued case-building into 2026, follows the same model as Sky ECC and EncroChat before it.
I don’t have a strong view on how much of the leak came from platform compromise versus SIM swap and device seizure. What I do know is that the operational security of the average directory-spoofing crew, judged by the 2026 case records, is not high. Shared hosting, shared profiles, shared kits, and shared comms platforms. Every layer of shared infrastructure is a subpoena target.
The LockBit and KillSecurity lesson for the roundup
Law enforcement roundups habitually inflate the arrest count by treating different campaigns as different actors. The threat intelligence data says otherwise. KillSecurity ran a leak site from mid-2024 until early 2025, then went quiet. In August 2025, a new site appeared as KillSec3, listed by trackers as the next iteration of the same operation. On 10 October 2025, KillSec3 posted 169 victims. Fifty-one of those matched victims were already listed under the old KillSecurity name months earlier.
That is one gang moving house and bringing its old address book along, not two rival crews. LockBit 5.0 launched in December 2025, nearly two years after Operation Cronos, the February 2024 takedown that hobbled the group, and needed to look busy fast. Its biggest posting days arrive as bursts, not trickles. Thirty-four victims added on one occasion, 62 on another.
Any 2026 arrest tally that counts these rebrands as fresh organizations is overstating the disruption. The same operators keep showing up under new labels because enforcement rarely removes the human capital, only the infrastructure and the branding.
What the pattern actually predicts
Expect the next round of arrests to follow the redirect chain. The infrastructure providers, the Cloudflare Turnstile gate operators, the anti-detect profile vendors, and the Railway-style cloud hosting accounts are where the evidence lives. The operators who bought Profile X and routed through Redirect Y are downstream of that, and the order in which the arrests land reflects it.
For the privacy-conscious analyst, the takeaway is that directory spoofing is a shared-services problem cleverly disguised as a personal-security problem. The operator who built a perfect browser fingerprint still fails if their profile came from the same vendor as forty other operators on the same Railway IP block. The 2026 roundup is the record of that leverage being used.
This article is for research and threat-intelligence awareness only. Do not attempt to access, operate, or replicate any of the infrastructure discussed here.