Wireshark-dev: [Wireshark-dev] Re: Heuristic dissectors default on/off - selection?

From: Anders Broman <a.broman58@xxxxxxxxx>
Date: Thu, 20 Nov 2025 22:56:53 +0100
>Wireshark should behave like it always did

So we're done  nothinhg to be added to the program. ;-)

We have preferences to to tune wireshark to individual needs. The default should be someting resonable. I thought I found "rare" protocols very few people would find in their traces and the ones that would, would find them missing and enable the heuristic.

But it seems the majority dissagrees.



Den tors 20 nov. 2025 22:35Roland Knall <rknall@xxxxxxxxx> skrev:
Bad example, that protocol is actually quite in use in games and some industrial applications ;-)

Am Do., 20. Nov. 2025 um 20:48 Uhr schrieb John Thacker <johnthacker@xxxxxxxxx>:
Another group is "obsolete." I think even people skeptical about the idea in general are easily on board with the idea of disabling the Yahoo Messenger protocol that hasn't been a commercial protocol in well over a decade.

On Thu, Nov 20, 2025, 2:46 PM Triton Circonflexe <triton+enuiqr@xxxxxxxxxx> wrote:
The profile-based presets looks like a good approach.
How would these profiles get generated?
- Hard-coded lists?
- “Tags” in the dissectors indicating to which categories they belong?

In any case, we can start with a few obvious sets like the “safe” one proposed by John and most of the ones proposed by Anders (also not sure about Bittorrent as a category, seems too specific).
I may suggest the "Web" category including the dissectors for the content of the data since there’s not much heuristics between frame and HTTP.


Le mer. 19 nov. 2025 à 21:46, Anders Broman <a.broman58@xxxxxxxxx> a écrit :
Protocol groups might help. Should be at least x(10?) dissectors or large ones.
Group Ideas:
Telco ( Better name? POTS, 2G, 3g etc)
File Storage ( DCE-RPC etc)
Car industry (ITS, CAN? ...
HomeAutomation ( Zigbee? ...
Bittorrent?
Games
...
Best regards
Anders


Den ons 19 nov. 2025 kl 22:04 skrev John Thacker <johnthacker@xxxxxxxxx>:
On Wed, Nov 19, 2025 at 3:59 PM Anders Broman <a.broman58@xxxxxxxxx> wrote:
The problem as I see it is that even if we have good heurustic detection. Worst case we might try every heurustic against every packet in the trace and make no match. But if you have traces with say trift or suspected trift you can enable the trift heuristic. Now worst case is trying one heuristic for every packet.

Downside is you will have to know which heuristics to enable, otoh you can always enable all again.

There's a "No Reassembly" profile that is automatically generated by a Python scripts in the tools directory that disables all the reassembly related preferences. I think it would be helpful to have extra default profiles that target different levels of enabled heuristic dissectors. (A profile optimized for speed with very few enabled, only reliable ones, only ones you might see on the public Internet but not industrial protocols, etc.) I think that both inexperienced and experienced users alike might want to quickly switch between large numbers of heuristics enabled and disabled without having to do it individually. If I am trying to characterize a completely unknown capture where I don't know what is there I have a different use case than a network where I already have a good idea what to expect.

Cheers,
John
_______________________________________________
Wireshark-dev mailing list -- wireshark-dev@xxxxxxxxxxxxx
To unsubscribe send an email to wireshark-dev-leave@xxxxxxxxxxxxx
_______________________________________________
Wireshark-dev mailing list -- wireshark-dev@xxxxxxxxxxxxx
To unsubscribe send an email to wireshark-dev-leave@xxxxxxxxxxxxx
_______________________________________________
Wireshark-dev mailing list -- wireshark-dev@xxxxxxxxxxxxx
To unsubscribe send an email to wireshark-dev-leave@xxxxxxxxxxxxx
_______________________________________________
Wireshark-dev mailing list -- wireshark-dev@xxxxxxxxxxxxx
To unsubscribe send an email to wireshark-dev-leave@xxxxxxxxxxxxx
_______________________________________________
Wireshark-dev mailing list -- wireshark-dev@xxxxxxxxxxxxx
To unsubscribe send an email to wireshark-dev-leave@xxxxxxxxxxxxx