DEV Community

Faceless Satine
Faceless Satine

Posted on

We tested 62,541 free proxies from GitHub. Only 4% actually work.

If you've ever grabbed a "free proxy list" off GitHub, you already know the feeling: you paste 10,000 IPs into your scraper, and approximately none of them connect. I wanted to know exactly how bad it is, so I measured it.

Short version: of 62,541 free proxies pulled from public lists, only 2,236 (4%) were alive. And "alive" is generous — it just means the TCP handshake completed once.

How I measured it

I built a liveness checker that:

  1. Ingests the big public free-proxy lists on GitHub (the monosans/proxy-list-style aggregators — HTTP, HTTPS, SOCKS4, SOCKS5).
  2. Dedupes and re-checks every proxy through a controlled endpoint on a rolling 5-minute cycle.
  3. Records success/failure, latency, and protocol over a 7-day window.

No sampling — the full set, continuously. A proxy counts as "alive" onlest in the most recent check.

The results

4% alive. 2,236 of 62,541. That's the number at any given moment; ttly.

Alive-rate by protocol:

Protocol Live now Alive rate
HTTP 1,474 6%
SOCKS5 479 2%
SOCKS4 247 2%
HTTPS 36 10%

HTTP dominates the live pool by raw count, but every protocol is brutal. HTTPS has the "best" alive-rate (10%) only because so few are listed to begin with.

It gets worse when you look at stability:

  • Median latency of live proxies: ~2,900 ms. Nearly 3 seconds per request — before your target server even responds.
  • Uptime across the live set: 44%. A proxy that answers now has a con the next check. So the usable pool at any instant is a fraction ofeven that 4%.

Put together: you might scrape a 60k list, find ~2k that ping, and be left with a few hundred that are actually usable for more than one request — each one adding ~3s of
latency.

Why free proxy lists are like this

  • They're recycled instantly. The moment an open proxy is discoverered by everyone until it dies or gets firewalled.
  • No ownership = no maintenance. Most are misconfigured servers or honeypots, not intentional services. Nobody keeps them up.
  • Churn is the whole game. By the time a list is committed to GitHud. Lists that "update daily" are mostly re-listing corpses.

The security part nobody mentions

Free proxies aren't just unreliable — some are actively harvesting. An nd your target and sees everything not wrapped in TLS. Treat any traffic through a random free proxy as public. Never send credentials or anything sensitive through one.

What this is actually good for

Free proxies do have a narrow, legit use: throwaway, low-stakes, high-tolerance-for-failure tasks where you can retry infinitely and don't care about latency or getting a
clean IP. Building a hobby scraper to learn on? Fine. Anything you depeill eat you alive.

Data + methodology

Full breakdown — alive-rate by protocol, geography, latency distributiop — is published here and refreshed continuously:

👉 **[Free Proxy Statistics report](https://proxmint.com/free-proxies/

Disclosure: I run Proxmint, a paid proxy servibuilding a free proxy-liveness tool, and the numbers were bleak-worthyenough to write up. The report page is free to read and cite — raw figures on request.

What's your experience been with free proxy lists? Curious if anyone's seen a source with a meaningfully better hit-rate.

webscraping #proxy #opensource #data

Top comments (0)