too many packets lost ...

[kings-way]

Have you noticed that the packet loss rate is a little high?
Many ICMP (ping) packets between station and the fakeap may be lost, for me, more than 50% sometimes. And I am sure it's not the wifi-adapter's problem.
If I use airbase-ng to setup an AP, then the situation will be much better.

I am not an expert in the network field.
So, I am just curious about that whether it is the problem of scapy or the way you handle dot11 packets?

[rpp0]

You are right; unfortunately Scapy is inherently slow at processing packets. This is because it will try to dissect each frame. If Scapy can't keep up, it will drop frames. I tried to mitigate the issue with the following line:

TCP.payload_guess = []

This removes the payload dissector for layers higher than TCP. You could try to do the same with the IP layer for slightly less overhead -- if you only need ICMP replies. So:

IP.payload_guess = []

Airbase-ng is written in C, uses libpcap, and does not automatically dissect, which makes it much faster at processing packets. If you require performance, libpcap is the way to go. If you quickly want to prototype or attack a Wi-Fi protocol and performance is not crucial, then scapy-fakeap will be more convenient :).

[kings-way]

Thanks for replying!
Actually, I am writing a tool making use of station probe packets to dynamically setup AP with the same SSID in probe packets. Yeah, it's a tool for pentest.
So, I have to provide stable Internet access for the stations.
Additionally, I translate dot11 frame to or from ethernet frame in my code. I think 'IP.payload_guess' may make some sense for me. I'll try it later.
Thanks again.

[rpp0]

You're welcome, sounds like a cool project too! I did the same thing a while ago with hostapd. Hostapd works by putting your interface in AP / master mode (which has the advantage of automatically responding to ACKs). Might be worth taking a look at as an alternative to airbase-ng.