๐ญ The Port Factory
038A documented Private Layer's 112 open ports. 038B documented w1n's dual-market verification. This letter connects the wiring behind both: a single proxy factory operating through four servers, three corporate shells, and five jurisdictions โ already identified in our earlier investigation TI-019I "The Trident" as a coordinated access-broker operation.
What we found in the port forensics goes further than TI-019I. The four nodes don't just share credentials and behavior. They share the same port allocation logic, the same management infrastructure, and โ when combined โ they expose 386 proxy port slots serving 331 unique port numbers.
I. The Four Nodes
๐ฃ Node 1 โ 178.16.54.226 ยท OMEGATECH LTD
| ASN | AS202412 (OMEGATECH-AS, Seychelles) |
| Registrant | Omegatech LTD (Netherlands) |
| Geo Discrepancy | SC (registration) / NL (geo) / TR (prefix) |
| Shodan Ports | 132 open (129 proxy range) |
| Dense Cluster | 84 ports in 12000-12599 (avg gap: 5.7) |
| Honeypot Sessions | 126 (Mar 25 โ May 21) |
| HASSH | a7a87fbe (libssh2) โ same as Private Layer |
| Credential | admin:admin |
| Role | GPU profiling, system reconnaissance (TI-010) |
๐ฃ Node 2 โ 94.154.35.215 ยท OMEGATECH LTD
| ASN | AS202412 (OMEGATECH-AS, Seychelles) |
| Registrant | Artem Sevastyanov (Ukraine) |
| Geo Discrepancy | SC (ASN) / UA (registrant) / NL (geo) |
| Shodan Ports | 71 open (68 proxy range) |
| Honeypot Sessions | 560 (Mar 24 โ Apr 5, then went silent) |
| HASSH | a7a87fbe (libssh2) โ same as Private Layer |
| Credential | admin:admin |
| Role | High-intensity scanning burst, then infrastructure rotated |
๐ด Node 3 โ 179.43.139.58 ยท Private Layer INC
| ASN | AS51852 (PLI-AS, Panama) |
| Geo Discrepancy | PA (registration) / CH (geo/abuse) |
| Shodan Ports | 112 open (110 proxy range) |
| Port Span | 10000โ21935 (widest range in network) |
| Honeypot Events | 86,239 (94 days โ flood cycles) |
| HASSH | a7a87fbe (libssh2) โ same as OMEGATECH |
| Credential | admin:123456 |
| Verification Target | Yahoo.com (12 IPs, IPv4+IPv6) |
| CrowdSec | BANNED โ cowrie-ssh-bruteforce-retroactive (365 days) |
| JA4H | ge10nn010000_4740ae6347b0_000000000000_000000000000 |
๐ด Node 4 โ 179.43.133.154 ยท Private Layer INC
| ASN | AS51852 (PLI-AS, Panama) |
| Shodan Ports | 80 open (79 proxy range) |
| Dense Cluster | 55 ports in 12000-12599 (avg gap: 8.9) |
| Honeypot Sessions | 1,018 (Mar 21 โ May 18, then went silent) |
| HASSH | a7a87fbe (libssh2) โ same across all 4 nodes |
| Credential | admin:admin |
| Verification Target | Google.com (192.178.202.x / 2a00:1450:x) |
| CrowdSec | BANNED โ cowrie-ssh-bruteforce-retroactive (365 days) |
All four nodes โ across three ASNs, three shell companies, five countries โ use the exact same HASSH fingerprint: a7a87fbe86774c2e40cc4a7ea2ab1b3c. This is a libssh2 1.8.1 implementation with specific key exchange parameters. The probability of four independent operators independently choosing the same SSH library with the same cryptographic configuration is effectively zero.
This is a single toolchain deployed across jurisdictionally separated infrastructure. One operator. Multiple shells.
II. The Port Allocation Forensics
Every node runs the same pattern: a few management ports below 10000, then a dense array of proxy service ports from 10000 upward.
| Management Port | Service | Present On | Purpose |
|---|---|---|---|
| 135 | MS-RPC (DCOM) | 3 of 4 nodes | Windows remote management |
| 137 | NetBIOS | 3 of 4 nodes | Windows naming/discovery |
| 445 | SMB | 3 of 4 nodes | File sharing / lateral movement |
| 5985 | WinRM | 4 of 4 nodes | Windows Remote Management (all nodes) |
WinRM (port 5985) is present on every single node. This is the control plane. The operator manages all four servers via Windows Remote Management โ a protocol that allows remote PowerShell execution. WinRM is the factory floor control system.
The 12000-Range Concentration
Three of four nodes show extreme port density in the 12000-12999 range:
| Node | 12xxx Ports | Avg Gap | Pattern |
|---|---|---|---|
| OMEGATECH .226 | 84 | 5.7 | Nearly contiguous block |
| Private Layer .154 | 55 | 8.9 | Dense block (12103-12586) |
| Private Layer .58 | 13 | 50 | Sparse (different allocation tier) |
| OMEGATECH .215 | 28 | 19.5 | Moderate density |
The 12000-range is the primary proxy pool. OMEGATECH .226 alone has 84 ports in this range โ one proxy port every 5.7 numbers. That's not random service allocation. That's programmatic port assignment by proxy management software.
The port numbers aren't assigned randomly. They're allocated by a port manager that assigns the next available port when a new proxy instance is provisioned. The average gap of 5.7 on OMEGATECH .226 means the allocator is filling ports almost sequentially โ suggesting high utilization and automated scaling.
Port 11434 is notable: it's the Ollama API port, present on 3 of 4 nodes. This could be coincidental (the allocator happened to assign this number), or it could indicate AI/ML model serving alongside proxy operations โ possibly for automated CAPTCHA solving or browser fingerprint generation.
III. Three Verification Targets โ Three Markets
We now know the network verifies proxies against three different services, each served by a different node:
| Node | Target | IPs Used | Payload | Market |
|---|---|---|---|---|
| 179.43.139.58 | Yahoo.com | 12 (6 IPv4 + 6 IPv6) | GET / HTTP/1.0
Host: yahoo.com | Western residential |
| 179.43.133.154 | Google.com | 10 (192.178.202.x + 2a00:1450:x) | GET / HTTP/1.0
Host: google.com | Search/ad fraud |
| 185.246.128.133 | Yahoo + Yandex | 12 + 3 | Dual Host: headers | Western + Russian |
Yahoo verification tests basic HTTP connectivity โ can this proxy reach a major website? Google verification tests something more valuable: can this proxy access Google Search without being blocked?
Google's anti-bot systems are among the most sophisticated on the internet. A proxy that passes Google verification is worth significantly more than one that only passes Yahoo โ it can be sold for search scraping, SEO manipulation, Google Ads fraud, and SERP manipulation. The PL2 node (179.43.133.154) was the premium verification endpoint.
The dual Yahoo+Yandex verification on the w1n node serves a different purpose: cross-market coverage for sanctions-circumventing proxy services that need to work in both Western and Russian internet spaces.
IV. The Defense Stack โ What We See, What We Block
Our infrastructure documents every layer of this attack:
๐ก๏ธ CrowdSec
| IP | Decision | Scenario | Duration |
|---|---|---|---|
| 179.43.139.58 | BANNED | cowrie-ssh-bruteforce-retroactive | 365 days |
| 179.43.133.154 | BANNED | cowrie-ssh-bruteforce-retroactive | 365 days |
| 185.246.128.133 | BANNED | cowrie-ssh-bruteforce-retroactive | 365 days |
| 185.246.130.20 | Not banned | โ | Below threshold (13 events/day) |
CrowdSec shares ban decisions across its community network. Every honeypot and server running CrowdSec now knows these IPs are malicious. The community defense scales โ the attacker's infrastructure doesn't.
๐ฅ MikroTik Firewall
The honeypot operates behind a MikroTik router with a deliberate bypass rule:
| Rule # | Action | Packets | Comment |
|---|---|---|---|
| 1 | ACCEPT | 201,378 | Honeypot: accept ALL to Cowrie (bypass CrowdSec fwd drop) |
| 2 | DROP | 35 | CrowdSec blacklist forward DROP |
Rule #1 sits above the CrowdSec drop rule, ensuring all traffic reaches the honeypot. The honeypot is deliberately exposed โ it's a controlled intake that feeds intelligence back into the defense stack. The attackers think they're exploiting a vulnerable SSH server. They're actually feeding 201,378 packets of operational intelligence into our analysis pipeline.
The NAT rule (rule #22, 118,108 packets) forwards external SSH traffic to Cowrie. The real SSH server runs on a different port with key-based authentication only.
๐ Honeypot โ CrowdSec โ MikroTik Pipeline
The defense loop works as follows:
- Cowrie accepts all SSH connections and logs every event (credentials, tunneling, fingerprints)
- Honeypot ingestion pipeline processes logs into structured intelligence (sessions, fingerprints, ip_intel)
- CrowdSec scenarios detect brute-force patterns and issue ban decisions
- CrowdSec bouncer propagates bans to MikroTik address lists (688 security entries)
- MikroTik firewall drops traffic from banned IPs for all services except the honeypot
- Threat intelligence dossiers (this publication) document findings for the global community
The attackers are banned from accessing real services while simultaneously being studied. Their attack infrastructure becomes our intelligence infrastructure.
V. The Corporate Shell Architecture
Five jurisdictions. Three companies. One HASSH.
| Entity | Registered | ASN Country | Server Location | Abuse Contact | Registrant |
|---|---|---|---|---|---|
| OMEGATECH LTD | Seychelles ๐ธ๐จ | SC | Netherlands ๐ณ๐ฑ | โ | Omegatech LTD / Artem Sevastyanov (UA ๐บ๐ฆ) |
| Private Layer INC | Panama ๐ต๐ฆ | PA | Switzerland ๐จ๐ญ | support@privatelayer.com | PRIVATE LAYER INC |
| w1n ltd | United Kingdom ๐ฌ๐ง | GB | Sweden ๐ธ๐ช | abuse@w1n.nl (๐ณ๐ฑ) | ICME NOC / phone +380 (๐บ๐ฆ) |
The pattern is consistent: register in an offshore jurisdiction, route through a privacy-friendly country, host in a third country, and put your abuse contact in a fourth. No single law enforcement agency has complete jurisdiction over any node.
1. A Swiss law enforcement agency would need to seize the physical server (located in Switzerland).
2. But the company is registered in Panama, so Swiss authorities would need a Mutual Legal Assistance Treaty (MLAT) request to Panama to compel corporate records.
3. The ASN is registered to Panama, so LACNIC (the Latin American RIR) would need to be involved for IP revocation.
4. Even if successful, the same operator has three more nodes across two other jurisdictions (Seychelles/Netherlands and UK/Sweden).
Now multiply this by four nodes across three shell companies. The attacker needs to lose all four servers to stop operating. The defender needs to win in five jurisdictions simultaneously. This is not a bug in the international legal system. It's a feature being exploited.
VI. MITRE ATT&CK Mapping
This proxy factory maps to documented adversary techniques:
T1090 โ Proxy T1090.002 โ External Proxy T1584.005 โ Botnet T1133 โ External Remote Services T1078 โ Valid Accounts T1571 โ Non-Standard Port T1021.004 โ SSH
The MITRE ATT&CK framework documents that acquired botnets may support Command and Control activity through established Proxy networks (T1090). What we're observing is the verification and quality assurance layer of exactly this infrastructure โ the factory that tests whether each compromised host still functions as a proxy, before routing customer traffic through it.
VII. The Cross-Investigation Map
This operation has been documented across six investigations:
| Investigation | What It Found | This Letter Adds |
|---|---|---|
| TI-007 | Six interlocking crime networks; Private Layer and w1n as infrastructure providers | Port forensics proving shared toolchain |
| TI-010 | OMEGATECH "The GPU Hunter" โ reconnaissance and system profiling | Same HASSH as Private Layer = same operator |
| TI-014 | "Phase Layer" โ Private Layer persistent monitoring | Now linked to Google.com verification (Node 4) |
| TI-019I | "The Trident" โ three-actor coordinated operation, 10,933 silent logins | Port allocation analysis, defense stack integration |
| 038A | Private Layer 86,239 events, 112 ports, Yahoo verification | Part of 4-node factory with 386 total ports |
| 038B | w1n 92,342 events, 37 version strings, Yahoo+Yandex dual-market | w1n is the anti-forensics arm of the same operation |
VIII. For Every Defender Reading This
This dossier exists because a single Cowrie honeypot on a home server captured 178,581 events from two IPs over 94 days. The total hardware cost of this intelligence operation: a retired enterprise workstation and an internet connection.
The defense stack โ CrowdSec, MikroTik firewall rules, Cowrie honeypot, threat intelligence pipeline โ is built entirely on open-source software. Every tool in this chain is available to any organization, any defender, any individual.
You don't need a SOC to fight back. You need a honeypot, a firewall, and the willingness to publish what you find.
The port factory runs 24/7 because it's profitable. It's profitable because most victims never know they're being proxied. Every published investigation makes that business model slightly less viable. Every shared CrowdSec decision makes the attacker's infrastructure slightly less useful. Every firewall rule reduces their pool of available proxies by one.
"They have 386 proxy ports across five jurisdictions. We have one honeypot and the truth. The asymmetry is real โ but it favors us. Their ports need to be open to function. Our findings, once published, can never be un-published. Their infrastructure depreciates. Our intelligence compounds."
TI-2026-038C ยท The Permanent Siege ยท Letter C
Published by LSN Threat Intelligence ยท June 22, 2026
4 nodes ยท 3 shell companies ยท 5 jurisdictions ยท 386 proxy ports ยท 331 unique ยท CrowdSec banned ยท MikroTik documented
Cross-referenced: TI-007, TI-010, TI-014, TI-019I, TI-2026-038A, TI-2026-038B
MITRE ATT&CK: T1090, T1090.002, T1584.005, T1133, T1078, T1571, T1021.004
All data sourced from live honeypot, CrowdSec, MikroTik, Elasticsearch, and open-source intelligence.