Forensic network dossier · Techloq Filter v2.5.2

How Chrome traffic reaches Techloq — and why it won't let go.

A plain-language record of the socket evidence: what was directly observed, what is only inferred, and why ten attempts to switch the filter off on this machine were refused at the kernel.

Evidence-based Windows · PowerShell October 2026 Owner's own device
Chrome network service
PID 17692
Confirmed
Techloq local listener
TCP 10001
Confirmed
Techloq process
tps.exe · 4612
Confirmed
Outbound endpoint
200.12.168.50:443
Confirmed

Disable attempts

10 methods
8 refused
2
Refused at the kernel · 8 Took effect, then reverted · 2

Finding confidence

7 findings
3 high
1
3
High · 3 Medium–high · 1 Unproven · 3
1

Case overview

The one-paragraph version, and the line between fact and inference this report keeps throughout.

The supplied TCP socket inventory shows that Chrome's dedicated network service process (chrome.exe, PID 17692) holds several established loopback connections to port 10001, which tps.exe (PID 4612) listens on. The same Techloq process simultaneously holds many established connections from the Wi-Fi address 192.168.68.60 out to 200.12.168.50:443.

That pairing is strong evidence of a local Techloq network intermediary sitting in the browser's path. It is not yet proof that a given Chrome request is forwarded to a given outbound socket, that all browsing passes through Techloq, or that HTTPS payloads are decrypted. Throughout this dossier, observed facts are separated from inference, and every conclusion carries a confidence level.

A second thread runs alongside the traffic analysis: between May and June 2026, ten distinct methods were tried to disable the filter on this machine without rebooting. Two succeeded briefly and were reverted at the owner's request; eight were refused outright by tpd.sys, a WHQL-signed kernel filter driver.

2

Traffic architecture

What the sockets, the process list, the device comparison, and the HTTPS probe actually showed.

Observed path

Chrome
network service
Techloq local
127.0.0.1:10001
Remote endpoint
200.12.168.50:443
Observed TCP connection Inferred forwarding (not per-request verified)

Socket evidence

The client process was identified from its command line, which marks it as Chrome's network utility rather than a tab or renderer:

PID 17692 · chrome.exe C:\Program Files\Google\Chrome\Application\chrome.exe --type=utility --utility-sub-type=network.mojom.NetworkService

Techloq's listener bound both IPv4 and IPv6 wildcards (0.0.0.0:10001 and [::]:10001), so it accepts on every local interface, though reachability from another machine still depends on the firewall. Chrome's connections went specifically to the loopback address. Two further loopback clients appeared on IPv6: PID 9656, and PID 5392, identified as LensService.exe (Poly Lens). Paired sockets recurred with near-consecutive local ports:

127.0.0.1:63741 → 127.0.0.1:10001 ESTABLISHED 17692 chrome 192.168.68.60:63742 → 200.12.168.50:443 ESTABLISHED 4612 tps.exe 127.0.0.1:61978 → 127.0.0.1:10001 ESTABLISHED 17692 chrome 192.168.68.60:61979 → 200.12.168.50:443 ESTABLISHED 4612 tps.exe 127.0.0.1:58233 → 127.0.0.1:10001 ESTABLISHED 17692 chrome 192.168.68.60:58234 → 200.12.168.50:443 ESTABLISHED 4612 tps.exe

Local and outbound ports opening one apart is consistent with one program creating both sockets in quick succession. It is circumstantial: it does not prove byte-for-byte forwarding or map any specific URL to a specific outbound socket.

On-device vs. off-device comparison

TestTechloq deviceSecond deviceReading
Wi-Fi / gateway192.168.68.60 / .68.1192.168.1.221 / .1.254Different networks — major confounder
DNS for filter.techloq.com18.130.77.2018.130.77.20Same destination resolved
TCP → 18.130.77.20:443successtimeoutDifference real; cause not established
TCP → 200.12.168.50:443successnot suppliedCannot compare this endpoint yet
HTTPS HEAD request200 OK (after revoke flag)timed out ~10 sOnly Techloq device reached a response
WinHTTP / env / user proxydirect; none; ProxyEnable=0sameNo conventional proxy configured
ICMP ping to AWS IPnot suppliedtimed outICMP block ≠ HTTPS block

Uncontrolled comparison. The two machines sat on different subnets, gateways and upstream routes, so none of these differences can be pinned on Techloq by itself. A clean test puts both devices on the same Wi-Fi and gateway before repeating every probe.

DNS, routing & HTTPS

Both machines resolved filter.techloq.com to 18.130.77.20, whose reverse name ec2-18-130-77-20.eu-west-2.compute.amazonaws.com places it in AWS London. A traceroute from the Techloq device began at the gateway, held near 30–34 ms through hops 10–12, rose to 104–138 ms at hops 13–14, then went silent from hop 15.

050100150 ms silent from hop 15 31104138 101112131415+ hop
Round-trip latency by hop on the Techloq device (approximate; hops 1–9 omitted). Latency roughly quadruples at hops 13–14, then responses stop.

A plain HTTPS request first failed with a Windows Schannel revocation error, then succeeded once revocation checking was allowed to be best-effort:

curl.exe -I -v --max-time 10 https://filter.techloq.com/ → CRYPT_E_NO_REVOCATION_CHECK (0x80092012) curl.exe -I -v --ssl-revoke-best-effort https://... Established connection from 127.0.0.1 port 50668 schannel: remote party requests renegotiation [twice] HTTP/1.1 200 OK Content-Type: text/html Content-Length: 2498 Last-Modified: Wed, 23 Sep 2026 10:22:59 GMT Cache-Control: max-age=3600, must-revalidate access-control-allow-origin: * Cache-Status: node2.env1.dc2.us.techloq.com;detail=no-cache

Two details stand out. The connection to a public address was reported with a loopback source (127.0.0.1), which is unusual for an ordinary routed socket and is worth correlating against the OS socket table. And the Cache-Status header exposes an internal-looking node name — the segments read like node, environment, datacenter and region labels, but names alone fix neither hosting nor whether the response came from a remote service or a local agent. The revocation flag tolerates an unavailable revocation list; it does not switch off certificate validation.

3

Disable attempts & kernel defenses

Ten methods on the owner's own machine: what took effect, what was refused, and the reason the refusals hold.

Took effect, then reverted

MethodTargetStatus
Stop & disable servicewtd.sysreverted 6/10
Block extension (Group Policy)Chrome / Edgereverted 6/10

Both were reverted at the owner's request on 10 June 2026; Techloq was fully restored.

Where the refusals happen

8 refused, by layer
Process & thread4
Service control1
Registry1
Filesystem minifilter1
Network (WFP)1

Each refused attempt grouped by the kernel mechanism that denied it.

Refused — eight methods

MethodLayerRefusal
SCM stop of tps.exeService controlNOT_STOPPABLE
taskkillProcessaccess denied
NtSuspendProcess / SuspendThreadThreadaccess denied
DebugActiveProcessDebugger attachaccess denied
System Informer (KProcessHacker)Thread freezeaccess denied
fltMC unload tpdMinifilterDO_NOT_DETACH
FwpmEngineOpen (delete WFP filter)Network filterERROR_NOT_SUPPORTED
Registry writes to service keysRegistryaccess denied

A hosts-file redirect of Techloq's servers was also tried; tps.exe kept connecting and the browser failed closed rather than open, so it blocked rather than bypassed.

Why the refusals hold

The protection is a single WHQL-signed driver, tpd.sys, enforcing four kernel mechanisms at once, so a bypass aimed at one is stopped by another before it lands:

  • Object callbacks (ObRegisterCallbacks) strip the rights a handle would need to open or kill the process.
  • Thread interception hooks the suspend and terminate path, so even a freeze from kernel-assisted tools is denied.
  • A minifilter on the filesystem refuses to detach (DO_NOT_DETACH) and guards the install.
  • WFP callouts own the network path; deleting the filters is not permitted.

A different network does not bypass it. Switching to a phone hotspot or any other connection does not get around Techloq. The filtering lives in a kernel driver on the device itself, in the WFP callout path, so it inspects traffic on whatever interface the machine uses before it ever leaves. Changing the upstream network only changes who carries the already-filtered traffic.

Untried routes

Both of these step outside the running OS, which is where the live protections lose their grip:

  1. Safe Mode with networking — boots a minimal driver set; third-party services generally do not start. A prepared script (SafeMode_Step1.ps1) clears the SafeBoot entries first. Untested; assessed high-likelihood.
  2. Offline registry edit via Windows PE — mount the disk from boot media and edit the service hives while tpd.sys is not running. Slower, but effectively guaranteed.
4

Endpoint register

Three address types surfaced and should not be conflated — a public site, a public endpoint the Techloq process actually used, and local addresses.

AddressRoleEvidence & limits
filter.techloq.com
→ 18.130.77.20
Public filter siteDNS A on both machines; AWS eu-west-2 (London). No tps.exe socket to this IP appeared in the snapshot.
200.12.168.50:443Outbound endpoint used by tps.exeMany established sockets from PID 4612. Server role, protocol and location unverified.
104.21.85.32:443Earlier endpointFrom a prior snapshot; falls in a Cloudflare-associated range. Host unknown.
127.0.0.1:10001
0.0.0.0 · [::]
Local Techloq serviceListener owned by tps.exe; Chrome connects here. Wildcard bind does not prove LAN-reachable.

Also seen on the machine but not shown to be in the traffic path: WindowsFilterAgentWPFClient.exe (the Techloq agent UI), plus pythonw.exe and node.exe tied to a local Argon bridge.

5

Confidence register

Every conclusion, rated by how far the supplied evidence actually carries it.

high
Chrome's network service connects to Techloq's local listener
Directly observed in the PID and socket output.
high
tps.exe holds outbound HTTPS sockets to 200.12.168.50
Directly observed, repeatedly, from PID 4612.
med–high
Techloq acts as a proxy-like intermediary
Strongly supported by topology and recurring paired sockets; not a verified request mapping.
unproven
Every Chrome request is relayed to 200.12.168.50
No per-request, timestamped socket correlation was captured.
unproven
Techloq decrypts HTTPS contents
No certificate-chain comparison or decrypted payload was supplied.
unproven
Techloq caused the second device's timeout
Different gateways and subnets prevent causal attribution.
high
tpd.sys refuses every no-reboot disable attempt
Eight independent methods each refused at the kernel.
6

Evidence gaps & next probes

The open questions, and the single most direct piece of evidence that would answer each.

next
Does Chrome use Techloq as a configured proxy?

Export chrome://policy and read ProxyMode, ProxyServer and ProxyPacUrl. Absence would not rule out transparent interception.

next
Is every request forwarded to 200.12.168.50?

Capture netstat -ano immediately before, during and after one request; match Chrome's ephemeral port to a Techloq outbound port by timestamp.

next
Is HTTPS being decrypted?

Compare the certificate issuer and chain on the Techloq device against a trusted device on the same network. A Techloq-owned issuer, or a failed pin, would settle it.

next
Is a WFP driver redirecting transparently?

Read-only inventory of installed filter drivers and WFP providers and callout layers.

next
Why does the second device time out?

Same-network A/B test; collect the missing 200.12.168.50 result from the second device.

next
Is port 10001 reachable from the LAN?

Authorized read-only firewall-rule check, then a controlled Test-NetConnection to port 10001 from another device on the same network.

Scope. This describes observed traffic and a plausible architecture on the owner's own machine. It is not a claim of a vulnerability or of unauthorized surveillance, and no part of Techloq was disabled, altered or bypassed to gather it.