sample_id string | carrier_name string | radio_technology string | country string | mtu_bytes int64 | tcp_mss_bytes int64 | tcp_window_size int64 | tcp_wscale int64 | initial_ttl int64 | df_bit bool | p0f_signature string | inferred_os string | ip_trust_score float64 | bot_detection_rate float64 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
PXM2-UK-EE-5G-001 | EE UK | 5G NR Sub-6 | United Kingdom | 1,420 | 1,380 | 65,535 | 8 | 64 | true | 4:64:0:1380:65535,8:mss,nop,ws,nop,nop,ts,sok:df:0 | Cellular Baseband / Darwin Dual-Stack | 99.4 | 0.02 |
PXM2-UK-EE-5G-002 | EE UK | 5G NR SA | United Kingdom | 1,420 | 1,380 | 65,535 | 8 | 64 | true | 4:64:0:1380:65535,8:mss,nop,ws,nop,nop,ts,sok:df:0 | Cellular Baseband / Darwin Dual-Stack | 99.6 | 0.01 |
PXM2-UK-VODAFONE-4G-001 | Vodafone UK | 4G LTE Cat-4 | United Kingdom | 1,440 | 1,400 | 65,535 | 7 | 64 | true | 4:64:0:1400:65535,7:mss,sok,ts,nop,ws:df:0 | Cellular Baseband / Android LTE | 98.8 | 0.03 |
PXM2-UK-THREE-5G-001 | Three UK | 5G NR | United Kingdom | 1,420 | 1,380 | 64,240 | 8 | 64 | true | 4:64:0:1380:64240,8:mss,nop,ws,sok,ts:df:0 | Mobile Gateway / Qualcomm X65 | 98.2 | 0.04 |
PXM2-UK-O2-4G-001 | O2 UK | 4G LTE | United Kingdom | 1,420 | 1,380 | 65,535 | 7 | 64 | true | 4:64:0:1380:65535,7:mss,nop,ws,nop,nop,ts,sok:df:0 | Cellular Mobile / Telefonica UK | 97.9 | 0.04 |
DATACENTER-AWS-001 | AWS EC2 (Squid Proxy) | Gigabit Fiber | United Kingdom | 1,500 | 1,460 | 29,200 | 7 | 64 | true | 4:64:0:1460:29200,7:mss,sok,ts,nop,ws:df:0 | Linux 5.15 Server (OS Mismatch Flag) | 12.1 | 0.92 |
RESIDENTIAL-VIRGIN-001 | Virgin Media Broadband | Cable DOCSIS 3.1 | United Kingdom | 1,500 | 1,460 | 64,240 | 8 | 128 | true | 4:128:0:1460:64240,8:mss,nop,ws,sok,ts:df:0 | Windows 10/11 Desktop | 84.5 | 0.21 |
Mobile Proxy Fingerprint Analyzer: TCP SYN & MTU Benchmark Dataset
A comprehensive empirical dataset comparing network-level passive OS fingerprint signatures (p0f v3, Akamai, Cloudflare, F5 Big-IP) across physical 4G/5G mobile proxy pools, residential broadband connections, and cloud datacenter proxy gateways.
Dataset Summary
This benchmark records network transport layer signatures during initial TCP three-way handshakes (SYN and SYN-ACK packets), measuring how Carrier Grade NAT (CGNAT) gateways and physical cellular modems alter packet parameters.
Monitored Telemetry & Metrics:
- Maximum Transmission Unit (MTU) & TCP Maximum Segment Size (MSS): Capturing cellular encapsulation clamping (e.g. 1420 bytes on 5G NR networks vs 1500 bytes on Ethernet).
- TCP Window Size & Window Scale Factor (WScale): Determining baseline buffer allocations across Linux, Windows, macOS, and cellular baseband stacks.
- TCP Options Ordering: Analyzing option sequence headers (MSS, SACK Permitted, Timestamps, NOP, WScale) utilized by deep-packet inspection (DPI) firewalls.
- Initial IP Time-To-Live (TTL) & Don't Fragment (DF) Flag: Detecting operating system hops and tunnel encapsulation.
- p0f v3 SYN Fingerprint Hash: Full p0f signature strings evaluating OS classification accuracy.
- Anti-Bot Anomaly Flags: Detecting discrepancies between browser User-Agent headers (Windows 11/macOS) and proxy egress stacks.
The Antidetect Consistency Problem
Anti-bot systems operate dual-layer verification:
- Client-Side (Application Layer): Evaluates Canvas, WebGL, WebRTC, and AudioContext via browser fingerprinting.
- Server-Side (Network Layer): Evaluates TCP SYN options, MTU clamping, and passive OS signatures via p0f.
When scraping teams deploy anti-detect browsers with residential or datacenter proxies, an OS mismatch anomaly frequently occurs: the User-Agent claims Windows 11, but the TCP SYN packet reflects a Linux datacenter kernel.
Hardware cellular gateways—such as UK EE 5G Modems—eliminate this risk because traffic originates through authentic carrier baseband hardware under CGNAT IP pools. For an in-depth architectural breakdown, explore the AdsPower Fingerprint Synergy engineering guide.
Dataset Schema
| Field | Type | Description |
|---|---|---|
sample_id |
string | Unique identifier for the benchmark probe |
carrier_name |
string | Cellular provider or infrastructure classification |
radio_technology |
string | Protocol type (5G NR, 4G LTE Cat-4, Fiber, Cable DOCSIS) |
country |
string | Geographic deployment region |
mtu_bytes |
int | Measured Path MTU |
tcp_mss_bytes |
int | Measured TCP Maximum Segment Size |
tcp_window_size |
int | Initial TCP Window Size in SYN packet |
tcp_wscale |
int | Window scale exponent factor |
initial_ttl |
int | IP header initial Time-To-Live |
df_bit |
boolean | Don't Fragment flag state |
p0f_signature |
string | Extracted p0f v3 raw signature string |
inferred_os |
string | Inferred operating system / kernel version |
ip_trust_score |
float | Aggregated IP reputation score (0.00 to 100.00%) |
bot_detection_rate |
float | Empirical probability of triggering bot challenge (0.00 to 1.00) |
Official Infrastructure & References
- Homepage & Platform: https://pxm2.io/
- Integration Framework: AdsPower Fingerprint Synergy
- Dedicated Cellular Modems: UK EE 5G Modems
- Free Sandbox Testing: 24h Free Trial
- Rotating Mobile Proxies: https://pxm2.io/proxy-types/rotating-mobile-proxies/
License & Citation
Licensed under the MIT License. Data curated by PXM2 Mobile Proxies.
- Downloads last month
- 130