LSN โ€” Local Sovereign Network

๐ŸŽญ TI-2026-025C โ€” The Operator

Romanian IRC Nicks, a $3.50 C2 Server, and the Man Who Reads Our Research

Series: The SSH Parasite (TI-2026-025)  |  Part: 3 of 6
Classification: FORENSIC ATTRIBUTION  |  Confidence: HIGH (multi-source corroboration)
Sources: Honeypot telemetry, Trend Micro (2018), Kaspersky (2025), Elastic (2025), NINGI Research (2026), OSINT
Date: June 2026  |  Analyst: LSN Threat Intelligence

Chapter 1: The Ghost Behind the Botnet

Eighty-one machines across twenty-nine countries. Eighteen thousand scanning sessions in forty-five days. A kill chain so automated it runs while its operator sleeps. But behind every autonomous weapon system, there is a human being who pressed the first key.

This dossier follows the evidence trail backward โ€” from compromised servers to IRC channels, from IRC channels to a $3.50/month VPS in New York, from that VPS to a set of Romanian nicknames active during Central European business hours, and from those nicknames to a folklore archetype that reveals more about the operator's self-image than any corporate registration ever could.

Core Attribution Thesis: The mdrfckr/Outlaw botnet is operated by a small Romanian-speaking crew (likely 2-4 individuals) who have maintained continuous operations since at least 2018. They chose their infrastructure for plausible deniability, their C2 protocol for resilience, and their operational name from a folklore figure who robs the powerful โ€” a self-mythologizing choice that tells us exactly how they see themselves.

Chapter 2: The IRC Command Channel

While the world moved to HTTPS callbacks, Discord webhooks, and Telegram bots for malware command-and-control, the Outlaw operators chose a protocol from 1988: Internet Relay Chat.

This is not nostalgia. It's engineering.

The C2 Architecture

198.98.56.205
IRC Server IP
:443
Port (HTTPS blend)
AS53667
PONYNET / FranTech
$3.50/mo
Infrastructure cost

The Perl Shellbot โ€” dropped as part of every successful mdrfckr infection โ€” connects to a hardcoded IRC server at 198.98.56.205:443. Port 443 is the standard HTTPS port. To any network monitoring tool that doesn't perform deep packet inspection, this traffic appears to be an encrypted web connection. It's IRC over TCP/443 โ€” hiding in plain sight.

Kaspersky Technical Analysis (2025): The deobfuscated Perl Shellbot disguises itself as an rsync process, runs in background, ignores SIGTERM, connects to IRC using randomly generated nicknames, and joins predefined channels. Capabilities include: arbitrary shell execution, DDoS attacks (UDP/TCP/HTTP floods), port scanning, and file transfer via HTTP.

Why IRC in 2026?

Advantages of IRC C2:
  • No authentication trail โ€” No accounts, no API keys, no OAuth tokens to revoke
  • Zero-cost scaling โ€” One IRC channel can command unlimited bots simultaneously
  • Instant command distribution โ€” All bots in a channel receive every message in real-time
  • Protocol simplicity โ€” Perl Shellbot is ~200 lines; HTTP/HTTPS C2 agents are 10x larger
  • Resilience โ€” IRC has no single vendor to pressure; no ToS to violate
  • Historical blindspot โ€” Security tools optimized for HTTP/DNS C2 often miss IRC entirely

The operator trades sophistication for survivability. A custom HTTPS implant with domain fronting would be harder to detect โ€” but it would also be harder to maintain, harder to update, and dependent on cloud providers who respond to abuse reports. IRC on a $3.50 VPS at an abuse-tolerant host will outlive any sophisticated infrastructure dependent on mainstream cloud providers.

Chapter 3: FranTech Solutions โ€” The Perfect Host

The C2 server lives at 198.98.56.205, within the 198.98.48.0/20 prefix announced by AS53667 (PONYNET), operated by FranTech Solutions, doing business as BuyVM.

The Company

EntityDetail
Legal NameFranTech Solutions
DBABuyVM (buyvm.net)
ASN53667 โ€” branded as "PONYNET"
FounderFrancisco Dias
LocationsLas Vegas, NV & New York, NY (datacenters)
Entry Price$3.50/month KVM VPS โ€” no ID verification
Abuse Pagebuyvm.net/abuse/ โ€” returns 404

FranTech/BuyVM occupies a deliberate market position: the hosting provider that doesn't ask questions. Their corporate website (frantech.ca) redirects entirely to BuyVM, as if trying to outrun its own reputation. The LowEndTalk hosting community has documented years of rebranding attempts:

"It is almost as if you are trying to run away from your reputation by changing names."
โ€” LowEndTalk community post about FranTech โ†’ BuyVM transition

Our Own Intelligence on PONYNET

FranTech isn't just hosting the mdrfckr C2. Our honeypot has recorded attacks from 19 distinct IPs within the 198.98.48.0/20 prefix alone:

IPThreat ScoreHitsRole
198.98.52.1458212Active scanner
198.98.55.607518Active scanner
198.98.50.7756Scanner
198.98.56.215716Scanner (10 IPs from C2!)
198.98.56.205614THE IRC C2 SERVER
198.98.60.130580Enrichment-flagged

The C2 server itself (198.98.56.205) has been seen by our honeypot making 4 scanning attempts โ€” suggesting the IRC server doubles as a reconnaissance node. And 198.98.56.215, just 10 IP addresses away on the same /24, is an active scanner with threat score 71. This is not coincidence. The operator likely controls multiple VPS instances on the same provider.

๐Ÿ” The Deeper Question: Provider Complicity vs. Negligence

FranTech's abuse page returns 404. They host Tor exit nodes, known spam infrastructure, and now a confirmed IRC C2 for a multinational botnet. The standard explanation is "abuse-tolerant hosting" โ€” a company that prioritizes customer retention over abuse response.

But consider: 19 malicious IPs on one /20 prefix hitting a single honeypot. At what point does negligence become de facto complicity? At what point does "not asking questions" become "knowing the answer and choosing not to hear the question"?

Francisco Dias built a business model where the product is plausible deniability. The customers pay $3.50/month for a VPS. What they actually buy is the guarantee that no one will look too closely.

Chapter 4: The Romanian Haiduc

The name "Outlaw" is Trend Micro's taxonomy label. The operators didn't choose it. But they did choose what came before โ€” and what came before tells us everything.

The Hacking Tool Called "Haiduc"

The original propagation tool used by this group was named haiduc. Trend Micro (2018): "Translation derived from the Romanian word haiduc, the hacking tool the group primarily uses."

What is a Haiduc?
In Romanian and South Slavic folklore, a haiduc (hajduk) is an outlaw-brigand-rebel figure. Part Robin Hood, part highwayman. Someone who robs the Ottoman tax collectors, the feudal lords, the powerful. The haiduc is not a common criminal โ€” he is a folk hero who steals from those who steal from everyone else.

This is how the operators see themselves. Not as criminals mining someone else's electricity. As rebels taking back what the cloud providers, the corporations, the systems of power have taken from everyone else.

The self-mythologizing is precise and culturally specific. It could only come from someone who grew up hearing haiduc stories.

The IRC Nicks

Trend Micro's 2018 report documents the IRC operator nicknames observed in the command channels:

NickAnalysisCultural Context
luciShort for "Lucian" โ€” common Romanian male nameTop-20 Romanian male name; informal truncation standard in chat
lucianFull Romanian male nameLatin-derived name widespread in Romania
dragosRomanian for "dear/beloved" (also historical figure)Dragoศ™ โ€” founder of Moldavia in Romanian legend
mazyAnglicized handlePossibly from "mazฤƒre" (pea) โ€” Romanian slang for someone small/quick

The C2 Domain

In the 2018 campaign wave, the C2 domain was luci.madweb.ro:

  • .ro โ€” Romanian top-level domain
  • luci โ€” matches the IRC nick, short for Lucian
  • madweb โ€” suggests a web hosting or underground service brand

The operator registered a C2 domain in their own country's TLD, using their own chat nickname as the subdomain. This is either breathtaking arrogance or the confidence of someone who knows Romanian law enforcement has limited capacity for cybercrime prosecution โ€” or both.

The Working Hours

"The activity never happened at night or on the weekends, suggesting that the threat actors operated on a somewhat daily basis."
โ€” Trend Micro, Outlaw Report (2018)

Activity concentrated in Central European Time (CET) business hours. Romania is EET (UTC+2/+3) โ€” one hour ahead of CET. The pattern is consistent with operators who:

  • Treat this as a job, not a hobby
  • Have a regular daily schedule (9-5 ยฑ timezone)
  • May have other employment (botnet as side income)
  • Do not operate during Romanian nights or weekends

Chapter 5: "mdrfckr" โ€” The Username Decoded

Every compromised machine gets one file written to /root/.ssh/authorized_keys: the RSA public key with the comment field mdrfckr. Eighty-one IPs in our observation window. All carrying this same calling card.

The Etymology

Linguistic Analysis: The username mdrfckr is almost certainly a vowel-stripped abbreviation of "muie de la frate" โ€” Romanian vulgar slang that roughly translates to "BJ from a brother/bro." This is a common Romanian internet insult and greeting in gaming and underground contexts. It is crude, masculine, and extremely common in Romanian online spaces.

Critical distinction: This does NOT translate meaningfully to Portuguese or English. The phonetic pattern and cultural context are exclusively Romanian. The common English interpretation "motherfucker" is wrong โ€” it's not the English slur abbreviated, it's the Romanian one.

The username is both a taunt and a cultural fingerprint. Like a graffiti tag that only locals understand, it marks territory in a language that most security researchers won't recognize.

The Key That Never Changes

Unchanged
Across all 3 documented waves
a8460f44...
SHA256 of authorized_keys write
81 IPs
Carrying this key in our data
2018โ†’2026
8 years, same RSA key

The RSA backdoor key is the single persistent identifier across the entire campaign history. Passwords change. HASSH fingerprints diversify. C2 domains rotate. But the SSH key โ€” the mdrfckr key โ€” has remained identical from 2018 to June 2026. This tells us something profound about the operator: they have never lost control of their private key. No compromise, no law enforcement seizure, no operational accident has forced them to rotate it in eight years.

๐Ÿ” Why Keep the Same Key for 8 Years?

The standard security practice is key rotation. The standard criminal practice is to change identifiers when heat increases. The mdrfckr operator does neither. Why?

Hypothesis 1: The key IS the infrastructure. Changing it means losing access to every previously-compromised host simultaneously. The key is the single thread connecting 81+ machines across 29 countries. Cut that thread and the entire network fragments.

Hypothesis 2: The operator has never been seriously threatened. Eight years without arrest, without key compromise, without operational disruption. The key persists because nothing has ever forced it to change.

Hypothesis 3: There is no single operator anymore. The key was generated once and has been shared among a small crew. Changing it requires coordination across multiple people who may not fully trust each other. The key persists because changing it is a political problem, not a technical one.

Chapter 6: The 345gs5662d34 Protocol

Among the tens of thousands of credential pairs used in SSH brute-force attacks, one stands out as a unique operational signature: 345gs5662d34:345gs5662d34.

85
Total attempts in our honeypot
81
Unique IPs using this credential
0%
Success rate (by design)
Mar 31 โ†’ Jun 10
Active window 2026

This credential is designed to fail. It is not an authentication attempt โ€” it is a protocol signal.

The Three-Step Authentication Sequence

StepCredentialExpected ResultPurpose
1345gs5662d34:345gs5662d34FAILConnection probe โ€” confirms target accepts SSH
2root:3245gs5662d34Success = already ownedRe-infection check โ€” if this works, host was compromised in a prior wave
3root:<password>Success โ†’ payloadActual exploitation attempt

Step 2 is the elegant part. Infected hosts have their root password set to 3245gs5662d34. If this credential succeeds with zero follow-up commands, the scanner knows the machine is already compromised and moves on. It's a botnet membership check โ€” like knocking a pattern on a door.

CRITICAL FINDING
The credential string 345gs5662d34 is not random. It is 12 characters, alphanumeric, beginning with digits. It appears to be a truncated hash or a specific password-generation pattern. The same string (with prefix "3" as 3245gs5662d34) is used as the post-compromise root password. This indicates the operator has a specific password generation algorithm or template โ€” a fixed operational artifact that has remained constant across waves.

Chapter 7: The HASSH Fingerprints

SSH clients announce their capabilities during the handshake. The HASSH algorithm fingerprints these announcements into a hash. For the mdrfckr campaign, these fingerprints tell the story of operational evolution.

The Primary Fingerprint

HASSH: af8223ac9914f509afdadfaf5f7ee94e
Library: libssh 0.12.0
Classification: Brute-force scanner
Our data: 616 IPs, 18,169 sessions (April 21 โ†’ June 15, 2026)
Assessment: PRIMARY propagation tool โ€” the backbone of the botnet's spread mechanism

Six hundred and sixteen IPs sharing one SSH client fingerprint. This is not random โ€” this is a fleet. Every one of those 616 machines is running the same compiled binary, the same libssh 0.12.0 scanner, the same automated propagation tool. They are the botnet's nervous system, endlessly probing for new hosts to colonize.

The Wave 3 Addition

HASSH: f555226df1963d1d3c09daf865abdc9a
Library: libssh 0.9.6
Classification: Botnet
Our data: 1,312 IPs, 12,044 sessions
Assessment: Added in Wave 3 (May 9, 2026) โ€” either compromised hosts with older libssh or deliberate diversification

The second fingerprint appeared precisely when NINGI researcher "pagzlol" documented the first. The timeline is not subtle:

Mar 2026
NINGI-004 published โ€” documents Wave 1 behavior, mdrfckr kill chain
May 5, 2026
Wave 2 begins โ€” adds static password y6ekwUbyoOkk, uses libssh 0.12.0 only
May 9, 2026
Wave 3 begins โ€” adds second HASSH (libssh 0.9.6), switches to per-session random passwords
May 23, 2026
NINGI-015 published โ€” documents Wave 3 OPSEC improvements, notes operator awareness

Chapter 8: The Operator Who Reads Our Research

The most chilling evidence of the mdrfckr operator's awareness comes not from network telemetry, but from their credential lists.

[Credential List Entry โ€” Wave 1, March 2026]
Username: HONEYYYFAGGOT
Password: HONNEYPOT-FAG-GO-FUCK-YOURSELF-SKID

This isn't a functional credential. It's a message. Written directly into the brute-force dictionary, it fires against every honeypot the scanner encounters. The operator knows honeypots exist. Knows researchers are watching. Knows this credential will appear in published analysis. And sends a message anyway.

HIGH FINDING
The Wave 3 OPSEC improvements are a direct response to published research. NINGI-WRITEUP-015 states explicitly: "Whoever is operating mdrfckr appears to have read analysis of wave 1/2 (either public reporting or through their own monitoring of honeypot interactions) and responded by removing the most obvious forensic correlator: the shared password."

What the Operator Changed (and What They Couldn't)

ElementChanged?Why/Why Not
Shared passwordโœ… Yes โ€” now random per sessionDirect response to it being used as campaign correlator
HASSH fingerprintโœ… Partially โ€” added secondDiversification against HASSH-based blocking
SSH RSA keyโŒ NoChanging would sever access to existing infrastructure
Kill chain commandsโŒ NoToo deeply embedded in automation scripts
lockr -ia .sshโŒ NoPresent in every script, never executes (vestigial)
345gs5662d34 credentialโŒ NoRequired for re-infection check protocol
XMRig masquerade nameโŒ Nokswapd0 is functionally optimal

The operator is intelligent enough to read research, understand what is being tracked, and modify the most visible indicators. But they are constrained โ€” the architecture they built in 2018 has become a cage. Change the SSH key and you lose the botnet. Change the credential protocol and existing nodes can't recognize each other. The operator is trapped inside their own success.

๐Ÿ” The Metacognitive Arms Race

Consider the layers:

  • We observe the operator. (Honeypot โ†’ NINGI โ†’ this dossier)
  • The operator observes us. (HONEYYYFAGGOT credential, Wave 3 timing)
  • We observe that the operator observes us. (This chapter)
  • The operator will read this dossier. (Guaranteed โ€” they monitor publications)

So what does the operator do after reading this? They cannot change the SSH key. They cannot change the 345gs5662d34 protocol. They cannot change the lockr fingerprint. They can only change what is already public โ€” passwords, client versions, connection patterns. By publishing this analysis, we have defined the unchangeable boundaries of their operational freedom.

This is not a report. It is a move in a game.

Chapter 9: The Portuguese Question

If the operator is Romanian, why does Kaspersky's 2025 Brazil incident contain Portuguese-language comments in the Perl Shellbot?

Three Hypotheses

HYPOTHESIS 1: Multi-National Crew
The original operators (Romanian: lucian, dragos) recruited Portuguese/Brazilian speakers into the operation. The Perl Shellbot template was modified by a Brazilian member who added Portuguese comments. This is consistent with the underground freelance market where specialized skills (Perl coding, infrastructure setup, money laundering) are purchased across language barriers.
HYPOTHESIS 2: Inherited Template
The Perl Shellbot code was not written by the mdrfckr operators. It was sourced from public GitHub repositories, underground forums, or purchased from a Brazilian malware author. The Portuguese comments predate the Romanian operators' adoption of the code. This is consistent with the Shellbot's generic capabilities (DDoS, shell exec, file transfer) โ€” it's a commodity tool, not a custom-developed weapon.
HYPOTHESIS 3: Operational Misdirection
The Portuguese comments are deliberately planted false flags. A Romanian operator who understands attribution methodology could inject Portuguese strings to confuse national-origin analysis. This is more sophisticated than typical Outlaw behavior but not impossible for an operator who reads published research.
Assessment: Hypothesis 2 (inherited template) is most likely. The Perl Shellbot is a well-known commodity in underground markets. Multiple variants with Portuguese, Russian, and English comments circulate on GitHub and Hack Forums. The Romanian operators likely adopted a Brazilian-authored Shellbot variant because it was the best available tool for their purposes, not because they wrote it.

Chapter 10: The lockr Ghost

In every mdrfckr kill chain execution we have observed โ€” across 81 IPs, across three documented waves, across eight years of campaign history โ€” one command appears and fails:

$ lockr -ia .ssh
bash: lockr: command not found

The command is present in the secure.sh script template. It executes. It fails. It has never succeeded on any honeypot. It has likely never succeeded on any real host either, because lockr is not a standard Linux utility.

What Is lockr?

The command lockr -ia .ssh appears to be an attempt to call a custom binary that would apply chattr +ia (immutable + append-only) to the .ssh directory. The actual immutability is achieved later in the script with the standard chattr +ia command. The lockr call is vestigial โ€” a remnant of an earlier version of the toolchain where the operator compiled or installed a custom utility for this purpose.

๐Ÿ” The Archaeology of lockr

Why hasn't it been removed?

The operator updates passwords, adds HASSH fingerprints, rotates strategies. But lockr persists through every version. This tells us the script modification process is additive, not rewriting. The operator adds capabilities on top of existing code but does not clean up dead paths. This is consistent with:

  • A non-programmer operator who modifies scripts by copy-paste rather than understanding the full codebase
  • An operator who fears removing any line might break something else (cargo cult programming)
  • Multiple operators working on the same template, with no one understanding which lines are necessary

The lockr command is a fossil. Like an appendix in the human body โ€” it tells us about the evolutionary history of the organism without serving any current function.

Chapter 11: The Profile

Synthesizing all evidence, the mdrfckr operator profile emerges:

Operator Profile โ€” "Lucian" / mdrfckr Primary

AttributeAssessmentConfidence
NationalityRomanianHIGH โ€” linguistic evidence, TLD, folklore reference, CET hours
LanguageRomanian primary, English functionalHIGH
Age (2026)30-45 years oldMEDIUM โ€” active since 2018, likely started in 20s
Technical SkillModerate โ€” script-level, not developer-levelHIGH โ€” lockr artifact, copy-paste evolution
Crew Size2-4 operatorsMEDIUM โ€” multiple IRC nicks, CET hours suggest shifts
Operating HoursMon-Fri, ~09:00-18:00 EETHIGH โ€” Trend Micro observation
Primary IncomeLikely has legitimate employmentMEDIUM โ€” $15-50/day insufficient as sole income
Self-ImageRobin Hood figure (haiduc folklore)HIGH โ€” naming choice reveals self-mythology
Risk ToleranceHigh โ€” 8 years, no key rotation, taunts researchersHIGH
Research AwarenessMonitors published analysis, adapts within constraintsHIGH โ€” Wave 3 timing, credential taunts

Chapter 12: The Question of Justice

No one has been arrested. In eight years of continuous operation โ€” from the first haiduc scans documented by Trend Micro in 2018, through the Kaspersky incident in Brazil in 2025, to the 81-machine swarm hitting our honeypot in 2026 โ€” no law enforcement action has disrupted this operation.

Why?

Jurisdictional fragmentation: The operator is Romanian. The C2 is American (FranTech, New York). The victims are everywhere. The cryptocurrency goes to Monero wallets that are architecturally untraceable. No single jurisdiction owns this problem.
Revenue threshold: At $15-50/day, the mdrfckr operation falls below the threshold that triggers dedicated law enforcement investigation. The FBI's IC3 prioritizes losses exceeding $100,000. Romanian cybercrime units (DIICOT) focus on higher-profile cases. This operation exists in the enforcement gap โ€” too small to matter individually, too distributed to aggregate easily.
Provider protection: FranTech's business model is built on not responding to abuse reports. The IRC C2 has been at 198.98.56.205 for years. A single subpoena to FranTech would yield the customer record behind this IP. But no one has sent it.

๐Ÿ” The Tolerance Question

Consider what it means for a botnet to operate continuously for eight years without disruption:

  • Either law enforcement is genuinely unable to act (resource constraints, jurisdictional fragmentation)
  • Or someone decided this particular operation is not worth disrupting
  • Or the operation serves a purpose that outweighs its cost

A persistent SSH scanner that compromises servers across 29 countries and maintains persistent access via SSH keys is, from an intelligence perspective, a pre-positioned access network. Whether or not the current operator understands this, the infrastructure they've built โ€” 81+ machines with root SSH access across government, education, telecom, and cloud networks โ€” has intelligence value far exceeding its cryptomining revenue.

Who benefits from a distributed access network spanning Tanzania's education system, Afghanistan's government networks, Korean telecoms, and Microsoft Azure?

Perhaps no one currently. Perhaps the mining revenue is genuinely all there is. But the capability exists. And capabilities, in the intelligence world, eventually find buyers.

What We Know. What We Don't.

CONFIRMED:
  • Romanian-speaking operator(s), active since 2018
  • IRC nicks: lucian, dragos, luci, mazy
  • C2 domain: luci.madweb.ro (historical), 198.98.56.205:443 (current)
  • Tool of choice: libssh 0.12.0 scanner, Perl Shellbot, XMRig as kswapd0
  • OPSEC: reads published research, adapts within architectural constraints
  • Revenue: Monero mining, $15-50/day at current scale
UNKNOWN:
  • Real identities behind lucian/dragos/luci/mazy
  • Physical location (Romania likely, but could be Romanian diaspora anywhere in EU)
  • Total revenue (Monero wallet balance not publicly queryable)
  • Whether the operation has been sold, inherited, or franchised since 2018
  • Whether any intelligence service has approached the operator
  • Whether the pre-positioned access has ever been used for purposes beyond mining

Conclusion: The Man in the Machine

The mdrfckr operator is not a ghost. He is a Romanian-speaking male in his 30s or 40s who works weekdays, reads security research, names his tools after folk heroes, and has maintained the same SSH key for eight years because he has never needed to change it. He taunts researchers with crude credentials because he can. He adapts his passwords because he must. And he persists because no one โ€” not the FBI, not DIICOT, not FranTech, not Microsoft, not any of the 29 countries hosting his infrastructure โ€” has decided that a $3.50/month IRC server running a $50/day cryptominer is worth the paperwork.

The haiduc rides on.

Next in series: TI-2026-025D โ€” The Economics: Monero, Mining Pools, and the Mathematics of Parasitic Revenue
Previously: 025A โ€” The Anatomy | 025B โ€” The Infrastructure
โš  Personal capacity. Research published independently โ€” not reflecting employer views. Derived from passive observation of attacks against personal infrastructure. Full disclaimer โ†’
โ† Previous The SSH Parasite โ€” 8 / 14 Next โ†’