EMULE

I must commend the AI here – I wouldn't have thought of asking you whether you installed Eddie from scratch or copied over the profile.
Most of the times those AIs catch mentions of one or two words in contexts of other words and hallucinate about the rest of the meaning, but this one was a good answer.

Okay, I'll listen to you.
You're more professional than me.😊😊😊

Hi Archaon1,
I'm glad I could help you.
Six months ago, I was just a newbie, but thanks to the community's help, I grew rapidly.
Now, I can finally help others too.
That's the meaning of a community.
Haha, it's really satisfying to help others.
🎉🎉🎉

I'll look into it, thank you

Emule,

I'm glad to see that this side of internet isn't dead!

Little question if i may : i used unlocker to get rid of the files. Windows was saying that i needed the authorisation of ThrustDLL or something like that.
How would you have granted yourself a higher authority...on your own pc?
(The classic windows "I bought you, I assembled you...I OWN YOUUUU")

Hi Emule,

Hard to follow someone with 35 posts but it worked out!
I was a bit suspicious when you asked about deleting system32's files but my VPN is working fine now ^^

Thanks a lot, I wouldn't have found it on my own.


For the data : i did the suppression of all the files and the reboot of the network

For the sake of so-called security, Windows' system permission design can sometimes be quite annoying.
When I installed Windows, I configured Huorong Security Software with the highest privileges, locking the core Windows processes from being tampered with.
For any operations requiring high privileges, I use Huorong Security Software's toolbox, which can solve the problem with a single click, saving me a lot of trouble.
 

Hi Archaon1,
I'm glad I could help you.
Six months ago, I was just a newbie, but thanks to the community's help, I grew rapidly.
Now, I can finally help others too.
That's the meaning of a community.
Haha, it's really satisfying to help others.
🎉🎉🎉

Hello!

After a year of using AirVPN I'm very happy with the product. Website has no bloat whatsoever and it's super easy to find what you are looking for. A huge plus goes out for having an active forum available! Much better option compared to social media idiocies. Also port forwarding has been executed greatly - many other VPN services miss that altogether but even those which support it can't match AirVPN's easy-to-use robust system. Config generator is a great plus too since I'm using both WireGuard app and WireSock depending on the situation and needs. Both run just fine and very few VPN's could match this level of usability.

I sometimes have dissapointing speeds with P2P, but usually a simple server change fixes it.

Overall very happy customer. Please have a beer AirVPN staff, you've deserved it!

Yes, the addition of the AmneziaWG protocol can solve the connection problems for most people at present, and I hope that AmneziaWG can be used for a long time. If in the future, when the existing protocol can no longer connect, I believe that AirVPN will add a new protocol to solve the connection problem. I will always believe in your technology and capabilities, and I believe that you will always let us breathe real internet. I will always love you, AirVPN. Keep it up!😘😘😘

Hello!

Not anymore, and even less in the near future. HTTP/3 is quickly spreading. Today, HTTP/3 is used by 36.5% of all the websites, including major web sites inside countries that enforce blocks against VPN. Furthemore, blocking UDP as such is no more realistic, not even in China, where UDP has become an instrumental protocol for many companies in any sector (video streaming, video conference, VoIP, marketing, social media marketing, regime propaganda and more), for regime aligned or regime owned activities.
 
In China you have a near 100% success rate and no shaping (apart from the normal shaping for anything outside China) with the current Amnezia "weak obfuscation" (no CPS) implementation, i.e. at the moment you don't even need QUIC mimicking (which is anyway available and very effective). Currently, bypassing blocks via UDP than via TCP is more efficient in China.
 
At the moment there is nothing more effective than mimicking QUIC with the signature / fingerprint of an existing web site that's not blocked, and you have this option right now. We see > 95% success rate, which is better than the success rates of SSH (not exceeding 75%), shadowsocks and XRay, V2Ray etc (but a lot faster!). The success rate is similar to any VPN protocol over HTTP/2, but, again, dramatically faster.
 
We're glad to know it. It is also very flexible. Thanks to CPS, you may mimic any transport layer protocol built on UDP, for example DNS, QUIC, SIP.

Kind regards
 

Traveling to Uzbekistan. I cannot connect to any AirVpn server anymore. It worked fine in USA.
Any protocol available in Eddie UI do not work.
Unfortunately I paid for the whole year of AirVPN service. Looks like the XRay or something similar is the way to go...

Please add more tunneling options! Day by day censorship around the world becomes worse and worse, OpenVPN and Wireguard is already completely blocked in many countries because it is easily detectable using DPI. Since the mid 2010's people out there have been working very hard on much more censorship-resilient solutions rather than simple-to-detect SSH and SSL-wrapping. The days of the games of privacy are over, now VPNs are more and more used by the people around the world to break the information bubbles created by their oppressive governments.

Here are some examples of the new tunneling methods that you could add to your services:
https://github.com/erebe/wstunnel
https://github.com/cbeuw/Cloak
https://github.com/v2fly/v2ray-core
https://github.com/XTLS/Xray-core

Hello!
 
We're very glad to announce that Eddie Android edition 4.0.0 Beta 1 is now available.
This is a major update: for the first time Eddie Android edition features AmneziaWG complete support.

Eddie Android edition is a fully integrated with AirVPN, free and open source client allowing comfortable connections to AirVPN servers and generic VPN servers offering compatible protocols. Eddie 4.0.0 aims primarily at adding, besides the already available OpenVPN and WireGuard, a thorough and comfortable AmneziaWG support. 

AmneziaWG is a free and open source fork of WireGuard by Amnezia inheriting the architectural simplicity and high performance of the original implementation, but eliminating the identifiable network signatures that make WireGuard easily detectable by Deep Packet Inspection (DPI) systems. It can operate in several different ways, including a fallback, "compatibility mode" with WireGuard featuring anyway various obfuscation techniques.
 
What's new in Eddie 4.0.0
  AmneziaWG support Amnezia WireGuard API updated OpenSSL, OpenVPN3-AirVPN and WireGuard libraries see the complete changelog below  
AmneziaWG overview
  From the official documentation: https://docs.amnezia.org/documentation/amnezia-wg AmneziaWG offers:
Dynamic Headers for All Packet Types (compatibility with WireGuard: YES)
During tunnel initialization, the library generates a set of random constants applied to each of the four WireGuard packet formats: Init, Response, Data, Under‑Load. These constants:
Replace predictable WireGuard packet identifiers; Shift offsets of Version/Type fields; Modify reserved bits. As a result, no two clients have identical headers, making it impossible to write a universal DPI rule.
 
Handshake Length Randomization (compatibility with WireGuard: NO)
In WireGuard, the Init packet is exactly 148 bytes, and the Response packet is exactly 92 bytes. AmneziaWG adds pseudorandom prefixes S1 and S2 (0-64 bytes by default):
len(init) = 148 + S1 len(resp) = 92  + S2 Offsets of the remaining fields are automatically adjusted, and MAC tags are recalculated accordingly. In order to keep backward compatibility with WireGuard, S1 and S2 must be set to 0. 
  Obfuscation Packets I1-I5 (Signature Chain) & CPS (Custom Protocol Signature) (compatibility with WireGuard: partial, with fallback)
Before initiating a "special" handshake (every 120 seconds), the client may send up to five different UDP packets fully described by the user in the CPS format. In this way AmneziaWG can mimic perfectly QUIC, DNS and other protocols adding powerful methods to circumvent blocks. QUIC is particularly interesting as HTTP/3 is built on it and currently, from Chrome and other compatible browsers, 50% of traffic to/from Google is QUIC traffic. Therefore, blocking QUIC may have major disruptions for any ISP.
 
Junk‑train (Jc) (compatibility with WireGuard: YES)
Immediately following the sequence of I-packets, a series Jc of pseudorandom packets with lengths varying between Jmin and Jmax is sent. These packets blur the timing and size profile of the session start, significantly complicating handshake detection.
 
Under‑Load Packet (compatibility with WireGuard: YES)
In WireGuard, a special keep-alive packet (“Under-Load”) is used to bypass NAT timeouts. AmneziaWG replaces its fixed header with a randomized one, the value of which can be set manually. This prevents DPI from filtering short ping packets, ensuring stable tunnel connections, especially on mobile networks.
 
 
How to use Eddie with AmneziaWG

To enable AmneziaWG mode, just tap the connection mode available in the main and other views. It will rotate between WireGuard, AmneziaWG and OpenVPN. Set it to AmneziaWG. In its default AmneziaWG mode, Eddie will use all the possible obfuscation, except protocol mimicking, that keeps WireGuard compatibility, thus allowing connections to AirVPN servers. The default settings choice was possible thanks to the invaluable support of persons living in countries where VPN blocks are widespread. Such settings have been tested as working and capable to bypass the current blocking methods in various countries. You may consider to modify them if they are ineffective to bypass "your" specific blocks.
  In Settings > Advanced, you will find, at the bottom of the page, a new "Custom Amnezia WG directives" item. By tapping it you will summon a dialog that will let you customize any possible AmneziaWG parameter.

You can maintain backward compatibility with WireGuard in the dialog WireGuard section, or enable the full AmneziaWG support in the Amnezia section, which is not compatible (at the moment) with AirVPN WireGuard servers. This mode will be mostly valuable in a not distant future, when AirVPN servers will start to support AmneziaWG natively. You may also enable QUIC or DNS mimicking for additional obfuscation efficacy. 

In order to maintain WireGuard backward compatibility, with or without QUIC or DNS mimicking, you must set:
S1 = S2 = 0
Hn ∈ {1, 2, 3, 4}
H1 ≠ H2 ≠ H3 ≠ H4

Furthermore, do not exceed the valid limit of the J parameters (anyway Eddie will not let you do it). In this preview version, Eddie's formal control of the input data is based on the following document. We strongly recommend you read it if you need to modify manually parameters:
https://github.com/amnezia-vpn/amneziawg-linux-kernel-module?tab=readme-ov-file#configuration


Please do not modify In parameters if you don't know exactly what you're doing. 

Eddie implements QUIC and DNS mimicking and random obfuscation packets for each specific "I" parameter (by using the corresponding "Generate" button). You can enable them with a tap on the proper buttons. You may mimic QUIC and DNS even to connect to WireGuard based servers.

When you enable QUIC mimicking and you maintain WireGuard backward compatibility, you add a powerful tool against blocks, because the first packets will be actual QUIC packets. AmneziaWG will fall back to WireGuard compatibility very soon. However, when DPI and SPI tools, and demultiplexers in general, identify the initial QUIC flow, most of them will be unable to detect a WireGuard flow for several minutes. This has been tested thoroughly with deep packet inspection on Linux and FreeBSD based machines by AirVPN staff.

Therefore, in different blocking scenarios the QUIC mimicking increases likelihood of successful block bypass. NOTE: the same does not happen with DNS mimicking. In this case DPI / SPI tools identify the stream initially as DNS, but are much quicker (just in a few dozens of packets) to identify the stream as WireGuard's, after the initial DNS identification.
 
If you decide to test, please report at your convenience any bug and problem in this thread. If possible generate a report from the app in a matter of seconds: by tapping the paper plane icon on the Log view bar rightmost side you will generate a full system report which will include both log and logcat and have it sent to our servers. Then you just need to send us the link the app shows you (open a ticket if you prefer to do it in private).  
Download link, checksum and changelog
https://eddie.website/repository/Android/4.0.0-Beta1/EddieAndroid-4.0.0-Beta-1.apk
 
This is a build debug package and side load is mandatory.
 
$ sha256sum EddieAndroid-4.0.0-Beta-1.apk 617269290a0406237646cc0885e5b10f3916252f89fe82ba9ccb947354980fcb EddieAndroid-4.0.0-Beta-1.apk
Changelog 4.0.0 (VC 37) - Release date: 26 November 2025 by ProMIND
Native Library
[ProMIND] updated to version 4.0.0, API 10 [ProMIND] added Amnezia WireGuard API [ProMIND] updated to OpenVPN-AirVPN 3.12 (20251126)
AirVPNUser.java
[ProMIND] getWireGuardProfile(): added Amnezia support

ConnectAirVPNServerFragment.java
[ProMIND] showConnectionInfo(): added AmneziaWG logo display [ProMIND] onCreateContextMenu(): added AmneziaWG items [ProMIND] onContextItemSelected(): added AmneziaWG items [ProMIND] added method loadVPNProfile()
ConnectVpnProfileFragment.java
[ProMIND] added Amnezia support
EddieLibraryResult.java
[ProMIND] added Amnezia WireGuard API
QuickConnectFragment.java
[ProMIND] onCreateView(): added AmneziaWG logo display [ProMIND] updateStatusBox(): added AmneziaWG logo display
SettingsActivity.java
[ProMIND] added "Custom AmneziaWG directives" setting
SettingsManager.java
[ProMIND] added Amnezia specific settings and methods
SupportTools.java [ProMIND] removed method getVPNProfile()
VPN.java
[ProMIND] added methods enableAmneziaWireGuard() and isWireGuardAmneziaEnabled()
VPNManager.java
[ProMIND] added method isWireGuardAmneziaEnabled()
VPNProfileDatabase.java
[ProMIND] added AMNEZIA type
WebViewerActivity.java
[ProMIND] EddieWebViewClient.shouldOverrideUrlLoading(): it now properly opens android asset files
WireGuardClient.java
[ProMIND] added WireGuard tunnel node to constructor  [ProMIND] added methods for generating Amnezia's junk settings
WireGuardTunnel.java
[ProMIND] added support for Amnezia WireGuard [ProMIND] added Mode enum [ProMIND] added tunnel node to constructor 
EddieLibrary.java
[ProMIND] added Amnezia WireGuard API
Kind regards & datalove
AirVPN Staff

Great!
Eddie finally supports AmneziaWG, and UDP finally has a masquerade protocol. Another protocol has been added to the list of protocols for bypassing China's Great Firewall.

These are the AmneziaWG parameters I use in China. This set of parameters can reliably bypass the GFW. Staff can take a look.
Jc = 20;
Jmin = 50;
Jmax = 1000;
S1 = 0;
S2 = 0;
H1 = 3;
H2 = 1;
H3 = 4;
H4 = 2;

Hello,
I am a user from Mainland China. Over long connections, both SSH and SSL get blocked and the connection is reset, so none of the Eddie-supported protocols can maintain a long-term connection. Only the AmneziaWG protocol can sustain long connections without speed degradation—and it’s even faster and more stable than standard WireGuard. However, switching servers with AmneziaWG is somewhat cumbersome.
I would like to ask: could future versions of Eddie support the AmneziaWG protocol? I believe it could completely replace standard WireGuard. This would be a huge benefit for AirVPN users behind the firewall. The AmneziaWG protocol’s source code is open-source on GitHub under the MIT license and supports free commercial use. I strongly hope Eddie can add support for AmneziaWG.
Thank you!

Hello!

We're sorry to inform you that due to sloppy support by the datacenter provider (Racklot) we have decommissioned the server Metallah. Metallah went down on June the 18th, 2025, because IP addresses were null-routed. After more than a month, in spite of various contacts and solicitations, Racklot still fails to restore the routing. Our patience is over and we're acting accordingly. This was the last server still not supporting IPv6 (again for the laziness and the sloppy behavior of Racklot), so we finally have IPv6 support on every and each server.

Kind regards
 

I see. Have a look at these..

Connection refused means it was able to connect but on your host nothing is listening on port 18992.

Well, it's the worst ISP in China.They shape everything.I suggest you should switch to other ISP such as China Telecom , China Unicom or China Mobile.You may find some cheap plans on the RedNote.But BE CAREFUL.
If you switch your ISP to I mentioned above, you could try these servers.
China Telecom: NL , DE , US SJC , US LAX
China Unicom: nearly all european servers (except IE , NO , SE) (I recommend NL , DE) , US servers in the west coast.
China Mobile: JP , SG servers
I recommend using OpenVPN (IP Entry 3 & UDP & Port 443),since WireGuard is too easy to be identified (run a Wireshark and you know what I mean).
If you have IPv6 connection,you should use them first (unless it's too slow for you).The GFW is less aggressive on IPv6 compared to IPv4

I wouldn't say that, seeing as the project seemingly recommends using its own Linux kernel module, so only specialized projects may pull this out-of-tree module and compile it into their kernels. On a standard router, maybe even if you flash it with specialized ROMs like OpenWrt, you may find Wireguard, but not AmneziaWG. It also seems to need its own forks of standard Wireguard tools which you probably won't find in some distribution families. Maybe Debian at some point, later Ubuntu, and maybe maybe Fedora. On Arch, it might surface on the AUR (or probably is), and on SuSE on the OBS. Red Hat will never adopt it, and if Enterprise is not really interested, you get into a situation where single developers, or a single group of devs, are maintaining something used commercially again. It is not sustainable; you'll never know if it'll still be there in 5 years, or if internal disputes won't force the project to be forked and developed under a different name. But standard Wireguard is developed by well-known researchers, right in the kernel, and garnered enough commercial interest that some consumer networking companies implemented it as a feature. Wireguard is sustainable. AmneziaWG is not. It will never replace standard Wireguard.
Besides, the aim of Wireguard is not privacy. And most people around the forums (by topics created in the forums, at least) use the VPN not for the privacy aspect but because they want to torrent. What they're looking for is performance. You don't need AmneziaWG for that.

What I'm concerned about is the relationship Wireguard <> AmneziaWG. AmneziaWG would have the obligation to behave in a way a standard Wireguard behaves if server and client differ. I don't know how Wireguard will react if those fixed parameters talked about in the docs are altered. Maybe it's not even a problem, since AmneziaWG clients can connect here normally. At least according to some threads it doesn't seem to be a problem. Dunno.

But, no, a replacement is highly unlikely, both generally and specifically on AirVPN.

Great suggestion!

Hello,  

I am a user from mainland China. During my usage, I've noticed that many servers with low load bandwidth usage are actually slower, such as those in Taiwan and Japan.  

I’d like to offer a suggestion to AirVPN. Personally, I believe the evaluation criteria for server quality should be based on CPU usage over a certain period, such as the average CPU usage over half an hour, rather than load bandwidth. I’ve frequently encountered handshake timeouts when connecting to "sulafat," even though the displayed load bandwidth isn’t high. I suspect this is most likely related to high CPU usage on the server, which causes key resolution timeouts.  

If I could connect to a server with relatively idle CPU resources, I think the connection quality would be much better.  

After all, speed is closely tied to protocol overhead. For example, if a 1 Gbit/s server is connected to 100 users using the WireGuard protocol, the server’s load bandwidth might reach up to 900 Mbit/s. However, if it’s connected to 100 users using SSL + OpenVPN TCP protocol, the load bandwidth might only be 400 Mbit/s. Clearly, the latter scenario places a higher burden on the server, yet the load bandwidth appears lower.  

Therefore, I believe servers with lower CPU usage offer better quality, rather than those with lower load bandwidth. Using average CPU usage as a metric seems more scientific to me.  

Does my point make sense? Does anyone agree with what I’m saying?

Thanks.

Hello,  

I am a user from mainland China. During my usage, I've noticed that many servers with low load bandwidth usage are actually slower, such as those in Taiwan and Japan.  

I’d like to offer a suggestion to AirVPN. Personally, I believe the evaluation criteria for server quality should be based on CPU usage over a certain period, such as the average CPU usage over half an hour, rather than load bandwidth. I’ve frequently encountered handshake timeouts when connecting to "sulafat," even though the displayed load bandwidth isn’t high. I suspect this is most likely related to high CPU usage on the server, which causes key resolution timeouts.  

If I could connect to a server with relatively idle CPU resources, I think the connection quality would be much better.  

After all, speed is closely tied to protocol overhead. For example, if a 1 Gbit/s server is connected to 100 users using the WireGuard protocol, the server’s load bandwidth might reach up to 900 Mbit/s. However, if it’s connected to 100 users using SSL + OpenVPN TCP protocol, the load bandwidth might only be 400 Mbit/s. Clearly, the latter scenario places a higher burden on the server, yet the load bandwidth appears lower.  

Therefore, I believe servers with lower CPU usage offer better quality, rather than those with lower load bandwidth. Using average CPU usage as a metric seems more scientific to me.  

Does my point make sense? Does anyone agree with what I’m saying?

Thanks.