You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
New attacker agents are always random agents in the MAL GUI #30
The simulator is not only meant for training agents using ML. The idea is that it could also be used to find attack paths using different types of heuristics. But I see your point that defining attacker policy doesn't make much sense for attacker training. It does make sense for defender training though.
Activity
mrkickling commentedon Dec 11, 2025
We could easily change it to nothing or PassiveAttacker, but it would be nicer to be able to set the policy on an attacker.
sandorstormen commentedon Dec 11, 2025
Nothing feels better as default
mrkickling commentedon Dec 11, 2025
The idea was that if you create a scenario and then run it in a simulator, something would happen at least.
So we should:
sandorstormen commentedon Dec 11, 2025
Who is going to do this? I use the scenarios for machine learning. I never have an attacker type.
mrkickling commentedon Dec 11, 2025
The simulator is not only meant for training agents using ML. The idea is that it could also be used to find attack paths using different types of heuristics. But I see your point that defining attacker policy doesn't make much sense for attacker training. It does make sense for defender training though.