LSN โ€” Local Sovereign Network

Ghost Machines โ€” A Forensic Investigation Series
Letter F: The Convergence
Where All Lines Meet

๐Ÿ•ธ๏ธ The Convergence

This is the last letter.

Across five previous letters, we dissected the SSH exploitation ecosystem piece by piece: the proxy hunters who recruit routers (A), the operators who race for the first ten seconds (B), the delivery networks that deploy malware (C), the statistical census of seven botnets (D), and the economy that sustains them all (E).

This letter does not introduce new data. It draws the lines between the data that already exists.

And when you draw all the lines, a single conclusion emerges โ€” one that the cybersecurity industry, cloud providers, regulators, and intelligence agencies all know but none of them will state:

The infrastructure of cybercrime and the infrastructure of legitimate technology are not parallel systems. They are the same system.

I. The Map โ€” What We Proved

037A โ€” The Proxy Hunters

ISAEV Igor (AS200730, Kazakhstan). 6 IPs. 4,336 hits. 91% success. Custom Go scanner with 1,247-char reconnaissance command. Partner: SOLDATOV Alexey (AS209702). Enabler: Proton66 OOO (AS198953, ASN-DROP). Shell companies: novahost.kz, hostup.kz. Proved: residential proxy recruitment is automated, industrial, and run through identifiable shell companies.

037B โ€” The First Ten Seconds

Seven operator profiles: Cataloger (16 IPs, Telegram/MikroTik/GSM check), mdrfckr (6 IPs, RSA-2048 backdoor since 2019), meow (2 IPs, Google Cloud Mirai), Ghost (3 IPs, anti-forensics), GPU Hunter (3 IPs, NVIDIA detection), Competitor Killer (4 IPs, rival elimination), Heartbeat (1 IP, 111ร— BMOK). Proved: post-compromise is not chaos โ€” it is a competitive market with specialized roles.

037C โ€” The Delivery Networks

w1n ltd (AS42237, UK/UA/SE): 5,959 hits. Sia Nano IT (AS42099, Latvia): whisper C2, zero abuse reports. Shared HASSH fingerprint (Paramiko). notwork-monitoring: 4 builds in 79 minutes. Silent SCP uploads: 6 IPs, same hash. Proved: malware delivery is a cross-border enterprise with scanning and C2 split across different jurisdictions.

037D โ€” The Botnet Census

11,026 sessions. 63 fingerprints. 2,190+ IPs. 84 countries. Seven campaigns. Cloud providers host 58% of botnet infrastructure. Tencent/Aceville: 97 IPs (23% of one botnet). Netherlands: 68.2% success rate. libssh monoculture: 2,303 IPs on one open-source library. Proved: the attack infrastructure runs on legitimate cloud platforms, not shadowy underground servers.

037E โ€” Follow the Money

Five revenue streams: proxy ($0.50-5/IP/mo), mining ($2-100/server/mo), credentials ($1-10K/listing), ransomware ($200K+ avg), data ($5-250/record). ISAEV ROI: 56x-566x. Cloud botnets: infinite ROI (zero cost). Total economy: $10B+/year. Alarum Technologies (NASDAQ: ALAR) โ†’ NetNut โ†’ Popa botnet. Proved: cybercrime is an industry with quantifiable unit economics and publicly traded participants.

037F โ€” The Convergence (This Letter)

Where all lines meet.

II. The Three Convergence Points

Across all five letters and our broader investigation series (50+ published dossiers), three convergence points emerge โ€” places where multiple, independently documented findings point to the same structural reality.

Convergence Point 1: Tencent/Aceville โ€” The Cloud-Crime Pipeline

Tencent's Singapore subsidiary, Aceville PteLtd, appears in our investigation series at least five times across independent methodologies:

InvestigationFindingMethod
TI-2026-031A "Invisible Dragon"Aceville PteLtd registered as RDAP org for hundreds of Tencent Cloud IPs involved in SSH attacksRDAP analysis
TI-2026-031B-E seriesTencent Cloud infrastructure consistently appears in attack campaigns against the honeypotHoneypot correlation
037D "Botnet Census"97 out of 421 IPs (23%) in the libssh_0.9.6 botnet registered to Aceville PteLtd / TencentHASSH campaign analysis
037D ASN DossierAS132203: 181 IPs observed, avg threat 74, max 100. Classification: CRITICAL. Bulletproof assessment: TRUEAutomated threat assessment
037D RDAP contactsqcloud_net_duty@tencent.com and abuse@tencent.com appear in 126+ IP registrations across 9 countriesRDAP aggregation

Tencent is not a bulletproof host. Tencent is a $450 billion market-cap technology company listed on the Hong Kong Stock Exchange. They operate WeChat, QQ, and one of the world's largest cloud platforms. And their cloud infrastructure โ€” specifically the Singapore subsidiary Aceville PteLtd, registered at 16 Collyer Quay โ€” is classified as CRITICAL threat level in our automated assessment, with a bulletproof assessment of TRUE.

PeeringDB lists them as "Tencent Global" with aliases "QQ/Wechat." Their AS132203 has 1,073 IPv4 prefixes. Our honeypot has observed 181 of their IPs engaged in malicious scanning. Their own RDAP contact emails โ€” the abuse emails listed by Tencent themselves โ€” are the contacts for botnet infrastructure.

Q: You're saying a company worth $450 billion is running a botnet.
A: No. We're saying a company worth $450 billion hosts a botnet, has been told about it, and has not removed it. The distinction matters legally. It is irrelevant forensically. The packets that hit our honeypot from 97 Tencent IPs do not carry a disclaimer saying "Tencent does not endorse this scanning activity." The victim's SSH daemon doesn't check the attacker's corporate responsibility statement before accepting the connection. Whether Tencent is negligent or complicit is a question for regulators. What the data proves is that they are present โ€” in volume, persistently, across multiple independent investigations.
Convergence Point 2: The Jurisdictional Arbitrage Pattern

Every malicious operation documented in this series uses the same jurisdictional playbook. The pattern appears in:

OperationRegistrationInfrastructureOperationsC2/Payout
ISAEV (037A)๐Ÿ‡ฐ๐Ÿ‡ฟ Kazakhstan (shell co)๐Ÿ‡ต๐Ÿ‡ฑ Poland (IP geo)๐Ÿ‡ท๐Ÿ‡บ Russia (RDAP)Unknown
w1n ltd (037C)๐Ÿ‡ฌ๐Ÿ‡ง UK (company)๐Ÿ‡ธ๐Ÿ‡ช Sweden (servers)๐Ÿ‡บ๐Ÿ‡ฆ Ukraine (phone)๐Ÿ‡ฑ๐Ÿ‡ป Latvia (C2)
Omegatech (037A)๐Ÿ‡ธ๐Ÿ‡จ Seychelles๐Ÿ‡ณ๐Ÿ‡ฑ Netherlands๐Ÿ‡บ๐Ÿ‡ฆ Ukraine (RDAP abuse)๐Ÿ‡ณ๐Ÿ‡ฑ Netherlands
Private Layer (037A)๐Ÿ‡ต๐Ÿ‡ฆ Panama๐Ÿ‡จ๐Ÿ‡ญ SwitzerlandGlobal (57+ proxy ports)๐Ÿ‡จ๐Ÿ‡ญ Switzerland
Tencent/Aceville (037D)๐Ÿ‡จ๐Ÿ‡ณ China (parent)๐Ÿ‡ธ๐Ÿ‡ฌ Singapore (subsidiary)9 countries (IP geo)๐Ÿ‡ญ๐Ÿ‡ฐ HK (stock exchange)
Proton66 (037A)๐Ÿ‡ท๐Ÿ‡บ Russia (OOO)๐Ÿ‡ท๐Ÿ‡บ RussiaGlobal๐Ÿ‡ท๐Ÿ‡บ Russia

The same pattern appears in our earlier investigations:

  • TI-2026-026 "Phantom ASN": shell company ASN registered through chain of Seychelles โ†’ Netherlands โ†’ Germany entities
  • TI-2026-031 "Invisible Dragon": Tencent China โ†’ Aceville Singapore โ†’ global IP space
  • TI-2026-032 "American Mirror": US hosting providers (ColoCrossing, SingleHop, Psychz) with abuse rates worse than Chinese ecosystem
  • TI-2026-036 "Romanian Corridor": Romania as a transit jurisdiction between Eastern European criminal infrastructure and Western European markets

The pattern is not coincidence. It is architecture. Every operation distributes its legal identity across jurisdictions that do not cooperate on cybercrime enforcement. Registration in one country, infrastructure in another, operations in a third, payout in a fourth. No single jurisdiction sees the complete picture. No single law enforcement agency has authority over all components.

Q: This is the same pattern used for money laundering, tax evasion, and sanctions avoidance. It's the offshore finance playbook applied to network infrastructure.
A: It is the same playbook because it is the same lawyers. Company formation in Seychelles, Panama, and British Virgin Islands is handled by the same corporate service providers whether the client is evading taxes, laundering money, or operating bulletproof hosting. The jurisdictional arbitrage is a service โ€” it is sold by professionals who don't ask what it's for. The difference between an offshore shell company hiding wealth and an offshore shell company hiding a scanner cluster is exactly nothing. The legal structure is identical. The corporate registration documents are filed at the same offices.
Convergence Point 3: The Credential Monoculture

The most damning finding in the entire series is also the simplest: two credentials account for 93.6% of all successful authentications.

CredentialSuccessesSource IPsTarget DeviceManufacturer Responsibility
admin:1234563,37111IoT/network devicesFactory default. Never changed.
admin:admin1,47083Universal IoT/adminFactory default. Never changed.
pi:raspberry21613Raspberry PiWell-known default. Changed in 2022.
root:root3837Dev/test serversHuman laziness.

Every botnet documented in this series โ€” every proxy hunter, every malware deployer, every credential harvester, every competitive killer โ€” feeds on the same vulnerability: devices shipped with factory default credentials that are never changed.

The entire $10 billion SSH exploitation economy documented in 037E runs on a vulnerability that could be eliminated by a single firmware update requiring users to set a unique password during setup.

The Raspberry Pi Foundation did exactly this in April 2022 โ€” Raspberry Pi OS no longer ships with a default pi:raspberry credential. And yet, four years later, 13 IPs in our dataset still successfully authenticate with pi:raspberry. The devices were deployed before the fix and will never be updated.

Q: If the fix is this simple, why hasn't it been implemented universally?
A: Because the people who profit from the vulnerability are not the people who suffer from it. The device manufacturer profits from the sale. The ISP profits from the subscription. The cloud provider profits from the botnet's server rental. The proxy market profits from the compromised device's bandwidth. The victim โ€” the owner of the router running admin:admin โ€” bears 100% of the cost: their bandwidth consumed, their IP address blacklisted, their network compromised. This is the textbook definition of an externality. And externalities are not resolved by markets. They are resolved by regulation. Regulation that does not exist.

III. The Entity Graph โ€” Who Appears Everywhere

When you overlay every investigation โ€” the 037 Ghost Machines series, the 036 Romanian Corridor, the 031 Invisible Dragon, the 032 American Mirror, the 026 Phantom ASN โ€” certain entities appear again and again. These are the structural constants of the SSH exploitation ecosystem.

EntityInvestigationsRoleStatus
Tencent / Aceville PteLtd (AS132203)031A-E, 037D, 037FCloud provider hosting 23% of observed botnetPublicly traded ($450B)
OVH SAS (AS16276)032A-H, 037DCheap VPS with slow abuse responsePublicly traded (โ‚ฌ2.5B)
DigitalOcean (AS14061)032A-H, 037DCloud VMs for scan-and-burn operationsPublicly traded ($3B)
Contabo GmbH (AS51167)032A-H, 037DGerman budget hosting, botnet infrastructurePrivate (Munich)
ColoCrossing / HostPapa (AS36352)032A, 037DUS hosting with highest abuse rate in our dataPrivate (Buffalo, NY)
Proton66 OOO (AS198953)037ABulletproof host, ASN-DROP, linked to PROSPERORussian OOO
Omegatech LTD (AS202412)037A, 037DSeychelles shell, Spamhaus DROP, 1,988 S/IPShell company
Private Layer INC (AS51852)037A, 037DPanama/Swiss, 57+ proxy portsShell company
ISAEV Igor (AS200730)037A, 037DKazakhstan shell, go_scanner, proxy recruitmentShell company
w1n ltd (AS42237)037C, 037DUK/UA/SE/NL/LV โ€” 5-country identity dispersionUK registered
Korea Telecom (AS4766)037DTelecom with 15+ botnet-enrolled IPsPublicly traded
Vietnam Posts & Telecom037D4.8 sessions/IP โ€” highest persistence ratioState-owned

The entity graph reveals a three-layer architecture that is structurally identical to the legitimate cloud computing stack:

Layer 1: Infrastructure (Cloud + Bulletproof)

Tencent, OVH, DigitalOcean, Contabo provide compute. Proton66, Omegatech, Private Layer provide persistence. Same function as AWS/Azure/GCP in the legitimate economy โ€” but with different abuse policies.

Layer 2: Platform (Botnets + Tools)

libssh 0.9.6/0.11.x provides the scanning framework. Paramiko provides the delivery framework. Custom Go scanners provide specialized capability. Same function as Docker/Kubernetes in the legitimate economy โ€” orchestration of distributed workloads.

Layer 3: Application (Operations + Revenue)

Access brokers, proxy networks, cryptominers, ransomware affiliates, data exfiltrators. Same function as SaaS products in the legitimate economy โ€” specialized services built on shared infrastructure and platforms.

IV. What the Honeypot Proved That the Industry Won't Say

The cybersecurity industry operates on a narrative: criminals are sophisticated adversaries using advanced techniques. Nation-states deploy zero-days. APT groups operate with military precision. The threat is external, technical, and requires expensive products to defend against.

Our honeypot disproves this narrative.

93.6%
of logins use 2 factory passwords
58%
of botnet IPs on legit cloud
$0
marginal cost for most attackers
0
zero-days used
0
advanced techniques observed
1
vulnerability: factory defaults

No zero-days. No advanced persistent threats. No supply chain attacks. No social engineering. No phishing. No exploit kits. No novel techniques of any kind.

Just admin:123456.

The entire 11,026-session, 84-country, seven-botnet, $10-billion-dollar SSH exploitation economy runs on a single class of vulnerability: devices with factory default credentials. The most expensive cybersecurity products in the world โ€” the EDR platforms, the SIEM systems, the threat intelligence feeds, the SOC-as-a-service offerings โ€” are deployed to defend against threats that amount, in practice, to someone trying admin:admin on port 22.

Q: If the problem is this simple, why does the cybersecurity industry exist?
A: Because the cybersecurity industry does not profit from solving the problem. It profits from the problem existing. If every IoT device required a unique password set during initial configuration, if cloud providers verified customer identity before issuing compute resources, if hosting providers were legally responsible for abuse originating from their networks โ€” the SSH exploitation economy would collapse. And with it, the justification for a significant portion of the $200 billion global cybersecurity market. The industry's incentive is not to solve the vulnerability. It is to detect, report, and sell protection against its consequences โ€” indefinitely.

V. The Questions That Remain

The forensic evidence in this series supports the following conclusions beyond reasonable doubt:

  1. SSH exploitation is an industrial operation with specialized roles, supply chains, and quantifiable unit economics.
  2. The infrastructure runs primarily on legitimate cloud platforms, not underground servers.
  3. The economic model produces near-infinite ROI by externalizing all costs to victims.
  4. The vulnerability is a single class: factory default credentials on connected devices.
  5. Jurisdictional arbitrage is the legal equivalent of bulletproof hosting โ€” it makes enforcement impossible by design.
  6. The entities that profit from the problem โ€” cloud providers, device manufacturers, hosting companies, telecom operators โ€” have no financial incentive to solve it.

What the evidence suggests but cannot confirm:

  1. The libssh2 ghost fleet (037D) โ€” 628 IPs mapping SSH without attempting login โ€” may be a state-sponsored intelligence collection operation. The infrastructure cost, the discipline of zero-interaction, and the geographic breadth (48 countries) suggest a resource level beyond criminal enterprise.
  2. The overlap between ISAEV and SOLDATOV (Kazakhstan shell companies, identical patterns, similar infrastructure) suggests they are fronts for the same operation โ€” possibly connected to larger Russian-speaking organized crime networks operating through Central Asian jurisdictions.
  3. The Tencent/Aceville pattern โ€” botnet infrastructure on a state-connected Chinese cloud platform โ€” raises questions about whether the line between Chinese state-sponsored scanning and "criminal" botnet activity is meaningful.
  4. The residential proxy market โ€” where publicly traded companies purchase bandwidth from botnets โ€” represents a potential point of regulatory intervention. If stock exchanges required disclosure of bandwidth sourcing, the Alarum/NetNut/Popa model collapses.
Q: You've documented a $10 billion criminal economy running on the infrastructure of publicly traded companies, funded by factory default passwords, protected by jurisdictional arbitrage, and enabled by an industry that profits from the problem's existence. What do you actually recommend?
A: Three things that would end 90% of the problem tomorrow:

1. Mandatory unique credentials. Legislation requiring all internet-connected devices to ship with unique, randomly generated credentials โ€” like the UK's Product Security and Telecommunications Infrastructure Act (2022). If every router required a unique password, admin:123456 stops working. The entire botnet recruitment model collapses.

2. Cloud identity verification. Cloud providers should require verified identity (government ID, credit card with 3D Secure, business registration) before provisioning compute resources. Free trial abuse โ€” the meow model, the scan-and-burn model โ€” becomes impossible when every VM is tied to a verified human.

3. Hosting provider liability. If hosting providers were legally responsible for persistent malicious traffic originating from their networks โ€” the way landlords are liable for nuisance โ€” the bulletproof hosting model dies. Omegatech, Private Layer, and Proton66 exist because the current legal framework lets them profit from abuse with zero consequence.

None of these require new technology. SSH is secure. Authentication is solved. Key-based access is provably safe. The problem was never technical. The problem is that the wrong people have the right incentives.

VI. Closing

Ghost Machines: The Complete Series

6
Letters Published
7
Botnets Documented
2,190+
Unique IPs Analyzed
84
Countries Mapped
16
Malware Families
12
Named Actors
$10B+
Annual Economy Estimated
1
Honeypot. One SSH Port.
"Every piece of evidence in this series was collected by a single Cowrie honeypot running on port 22 of a single IP address in a 44-square-meter apartment in Iaศ™i, Romania. The server is a refurbished Dell Precision Tower 7910 with 503 GB of RAM, running K3S with 51 namespaces. The total cloud cost of this investigation was zero dollars."

The machines were always there. We just started counting them.

TI-2026-037F ยท Ghost Machines Series ยท Letter F โ€” Final Letter
Published by LSN Threat Intelligence
Data: LSN Cowrie Honeypot ยท threat_intel DB ยท dossier_intel DB ยท OSINT Library
50+ published dossiers ยท 8,000+ analyzed IPs ยท 271K+ entity links
All data sourced from proprietary honeypot deployment and open-source intelligence.
Zero cloud cost. Zero corporate sponsors. Zero editorial oversight.
Just the evidence.

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 Ghost Machines โ€” 6 / 6 Next โ†’