EDITOR’S NOTE — Aug. 17, 2026: This article has been substantially updated to reflect Hunt.io’s current findings. The research originally described approximately 2,100 servers associated with Friend Hosting LTD as command-and-control infrastructure. Hunt.io subsequently revised that characterization after Friend Hosting reviewed the underlying data and raised questions about how Keitaro traffic-distribution software had been classified. The original finding is retained below as part of the chronology, but it is no longer presented as Hunt.io’s current conclusion.
A Hunt.io investigation into Eastern European hosting infrastructure has undergone significant revisions after one of the providers identified in the research challenged how thousands of detected servers had been classified.
The June report originally said a single Bulgarian provider accounted for more than half of approximately 3,900 command-and-control servers detected during a three-month study. That provider was Friend Hosting LTD, which Hunt.io initially associated with roughly 2,100 of those servers.
Hunt.io no longer makes that claim.
After sharing its underlying data with Friend Hosting, the researchers said they had failed to distinguish clearly enough between command-and-control infrastructure and deployments of Keitaro, a commercial traffic distribution system that can be used legitimately or abused by threat actors. Hunt.io said the presence of Keitaro demonstrates that the software is running on a network – not that the network itself is operating C2 infrastructure.
The revised Hunt.io report still identifies more than 3,900 “threat activity enabling” servers across 302 Eastern European infrastructure providers. It now says Friend Hosting hosts many Keitaro instances that can be abused and notes that the provider worked with the researchers to analyze potential abuse.
Friend Hosting CEO Pavlo Khodnevych said his company’s review did not substantiate the original characterization.
“Our review did not confirm the original conclusion in the Hunt.io report that Friend Hosting was hosting a concentration of C2 servers. We shared our findings with Hunt.io, and we’re pleased that the final report was revised to better distinguish traffic distribution infrastructure from command-and-control infrastructure.”
Khodnevych said he viewed the episode as an example of security research being tested and refined rather than simply accepted at face value.
“I believe this is exactly how security research should work. Researchers publish their findings, affected parties review the data, and when additional facts emerge, the conclusions are refined. That’s not a weakness of the process—it’s a strength.”
What Changed
Hunt.io’s original report, published June 24, examined infrastructure across Belarus, Bulgaria, the Czech Republic, Hungary, Poland, Moldova, Romania, Russia, Slovakia and Ukraine.
Its initial framing centered on what appeared to be an extraordinary concentration of C2 infrastructure at one Bulgarian hosting provider.
Friend Hosting disputed that characterization and requested the underlying technical data. Hunt.io subsequently acknowledged that its original report had conflated the presence of certain software with evidence of malicious activity and revised its terminology from “C2” to “threat activity enabling.”
Following a further data-sharing exercise with Friend Hosting, Hunt.io updated the report again on July 10. The researchers said their overall counts had not changed, but how those detections should be characterized had.
How the finding changed
The dispute began with a striking number.
In its original June report, Hunt.io said it had identified roughly 3,900 command-and-control servers across Eastern European hosting providers. About 2,100 of them, the researchers initially said, were associated with Friend Hosting LTD — more than half of the total.
That finding made Friend Hosting the centerpiece of Hunt.io’s report and Security Point Break’s original coverage.
Friend Hosting challenged the classification after learning of the research through Security Point Break. The company said it had not initially been provided with the IP addresses, observation timestamps, technical indicators or other underlying data it needed to independently assess Hunt.io’s conclusions.
“We have not been provided with the IP addresses, observation timestamps, technical indicators, or any other data that would allow us to verify the conclusion that Friend Hosting LTD’s infrastructure included approximately 2,100 servers classified by the authors as C2 infrastructure,” the company said at the time.
The disagreement centered largely on Keitaro, a commercial traffic distribution system. Threat actors have long abused Keitaro to route victims toward phishing pages, malware downloads and other malicious destinations. But running Keitaro is not, by itself, evidence that a server is functioning as command-and-control infrastructure.
Hunt.io subsequently agreed that its original report had failed to make that distinction clearly enough.
In its first formal revision, the company acknowledged that the report had “conflated the presence of certain software with malicious impact” and called that “a failure in our editorial review process.” Hunt.io apologized to readers and the organizations referenced in the research and replaced much of its original “C2” terminology with the broader phrase “threat activity enabling.”
The researchers also softened references to “malicious” activity to “potentially malicious,” reflecting the fact that some of the tools they detected can have legitimate, unwanted or malicious uses depending on context.
A Second Review
The research changed again after Hunt.io shared its underlying data with Friend Hosting.
Friend Hosting reviewed the information, identified a group of accounts it believed warranted further investigation and told Hunt.io it had strengthened internal abuse-detection measures as a result.
But the provider said the cases that raised credible concerns involved traffic-distribution infrastructure, not infrastructure it could identify as command-and-control servers.
Hunt.io’s July 10 update explicitly recognized that distinction.
“Keitaro is a commercial traffic distribution system,” the researchers wrote. “Its presence on a network tells you the software is there. It does not tell you the network is running C2.”
Hunt.io said its underlying observations had not disappeared.
“Our counts have not changed. How we describe them has,” the researchers wrote.
The revised report no longer identifies Friend Hosting as the provider behind the original concentration finding.
Friend Hosting CEO Pavlo Khodnevych said the company’s investigation did not confirm Hunt.io’s initial conclusion that Friend Hosting was hosting a large concentration of C2 servers.
“We actually first learned about the Hunt.io report through Security Point Break’s article,” Khodnevych said. “Instead of rushing to dismiss it, we decided to review the data ourselves. Our review did not confirm the original conclusion in the Hunt.io report that Friend Hosting was hosting a concentration of C2 servers.”
Khodnevych said Friend Hosting shared its findings with Hunt.io and welcomed the researchers’ decision to revise the report.
“I believe this is exactly how security research should work,” he said. “Researchers publish their findings, affected parties review the data, and when additional facts emerge, the conclusions are refined. That’s not a weakness of the process – it’s a strength.”
The broader findings remain
The revisions significantly change how Hunt.io’s Friend Hosting finding should be understood. They do not erase the broader infrastructure activity documented during the three-month investigation.
Hunt.io’s current report describes more than 3,900 “threat activity enabling” servers observed across 302 infrastructure providers in Belarus, Bulgaria, the Czech Republic, Hungary, Poland, Moldova, Romania, Russia, Slovakia and Ukraine.
The distinction is important because much of the infrastructure Hunt.io observed sits one layer removed from what victims actually encounter.
Rather than simply hosting phishing pages or exploit sites, many of the systems identified by the researchers provided traffic distribution, remote administration, post-exploitation or other infrastructure that can be used as part of malicious operations.
Keitaro was among the most frequently observed technologies.
A joint Infoblox and Confiant study published in March identified roughly 15,500 domains associated with malicious Keitaro deployments during a four-month period, highlighting how extensively traffic-distribution systems can be abused in phishing, investment scams and malware campaigns.
Hunt.io also observed offensive and remote-management frameworks including Cobalt Strike, Sliver and Tactical RMM across infrastructure in the region.
“The dominance of Keitaro across the malware family distribution reflects the region’s established role in traffic distribution and redirect infrastructure, a foundational layer for malvertising, phishing, and exploit kit operations,” Hunt.io researchers wrote.
“The simultaneous presence of Cobalt Strike, Sliver, and Tactical RMM indicates that Eastern European hosting serves both commodity criminal operations and more sophisticated post-exploitation campaigns sharing the same infrastructure layer.”
The researchers also identified infrastructure associated with Cloud Atlas, the long-running espionage operation. Kaspersky researchers reported in May that Cloud Atlas remained active into early 2026, targeting government and diplomatic entities in Russia and Belarus.
The broader concern extends well beyond any single provider.
Europol’s 2026 Internet Organised Crime Threat Assessment warned that cybercriminal groups increasingly rely on complex hosting arrangements to make infrastructure harder to trace and disrupt. Dutch authorities also dismantled a bulletproof-hosting network this year that authorities linked to cyberattacks, disinformation operations and Russian sanctions evasion.
Hunt.io’s research offers another view into that infrastructure layer — but its revisions also illustrate the difficulty of determining what observed software or network activity proves about the operator hosting it.
In Friend Hosting’s case, what began as one of the report’s most dramatic findings ultimately became a narrower conclusion: Hunt.io observed substantial traffic-distribution infrastructure on the provider’s network, some of which could be abused, but the researchers no longer characterize those observations as evidence that Friend Hosting hosted 2,100 C2 servers.
Editor’s note — Aug. 17, 2026: This article has been substantially updated to reflect Hunt.io’s current findings. The research originally described approximately 2,100 servers associated with Friend Hosting LTD as command-and-control infrastructure. Hunt.io subsequently revised that characterization after Friend Hosting reviewed the underlying data and raised questions about how Keitaro traffic-distribution software had been classified. The original finding is retained as part of the chronology but is no longer presented as Hunt.io’s current conclusion. Security Point Break contacted Friend Hosting prior to the original publication deadline; the company responded after publication and subsequently participated in Hunt.io’s review of the underlying data.