Reading a detection

The fields on a detection, what each one is for, and how to turn a log entry into the exact config change that stops it happening again.

Last updated

Every detection Raven files carries an Additional info panel. It names the setting that fired and the value that caused it, which together tell you the one line of config to change, so you fix the cause instead of guessing at thresholds.

Where the panel is

Open a ban on the Bans List and the panel is a tile in the detail view, under the identifiers, with one row per field and a copy button on each value. The same payload is printed as the Additional Info block on the Discord log card, so you can read it without opening the dashboard.

A detection carries its context there. A kick or ban an admin issued carries a short record instead, saying who issued it and from where (see below), and only a detection gets the automatic false-positive flow on that ban.

The fields and what each one is for

Every detection stamps Config Option, without exception. The rest appear only when they are relevant to the check that fired, so a payload with two rows and a payload with eight are both normal.

FieldWhat it isWhat to do with it
Config OptionThe detection that fired, named exactly as it is in the config.The first half of the path to the setting: Detections.<category>.<Config Option>.
TypeWhich sub-check inside the detection fired.Maps to Types.<Type>. Turning that one off leaves the rest of the detection running.
MethodWhich numbered method inside the check fired. Not shown for AntiResourcesInjection or AntiHookNatives.Narrows a report further. Usually read rather than configured, though a few detections do take a Methods map.
TriggerPlain words for what crossed the line, such as Deleted Too Fast or Near Other Player.The fastest way to tell a spam rule apart from its neighbours without counting rule numbers.
RuleWhich numbered spam rule fired, on the entity, damage and pyrotechnic spam protections.Points at one rule inside that detection's Rules map. On the entity spam protection the map sits under the entity type, so the path is Types.<Entity Type>.Rules.<Rule>.
ThresholdThe count the rule was armed with, on the damage spam protections.Tells you what the rule was set to without opening the config. Not stamped by every spam protection.
Time WindowThe window the rule counted inside, in seconds, beside Threshold on the damage spam protections.The other half of the same rule. Raise it and the rule tolerates the same count spread over longer. On the executor and vehicle take-over checks the same key is evidence rather than a setting.
ResourceThe resource the check was about, on the resource detections.On AntiResourcesInjection it is the name for WhitelistResources. The start and stop checks have no whitelist, so there it is evidence rather than a lever.
Entity TypePeds, Vehicles, Objects or Pickups.Which Types entry inside the entity detection to look at. The spam protection has three of them, so its Rules map sits under one. Pickups exists only on the spawn protection, which has no rules at all.
Entity Population TypeHow the engine classified the entity it refused, on the entity spawn protection.Evidence rather than a lever. There is no setting keyed on it.
Entity ScriptThe resource that created the entity, on the entity spawn protection.The name to put in WhitelistResources. This is the field that fixes a spawn false positive.
Entity ModelThe model involved, as a name or a hash. Entities Models is the plural form the spam protection uses, with a count beside each repeated model.The value for WhitelistModels, which accepts either form. On EntitiesModelProtection it is the entry to remove from Models, or the one to add when the Mode field says allowlist.
NativeThe native named on an AntiHookNatives report. Not every report carries one.Evidence rather than a lever. AntiHookNatives has nothing to tune beyond Enable and Action.
Event NameThe event that was blocked, with Event Resource naming where it was registered.The event goes in WhitelistEvents, or the resource in WhitelistResources.
Weapon NameThe weapon involved, with Weapon Hash beside it.The value for WhitelistWeapons, which accepts either form.
DistanceThe distance involved in the report, in metres.Read it before assuming a cheat: a garage script moving a player produces the same number a cheater does.
VictimsThe players the burst of damage landed on, named, on the damage spam and blocked damage checks.Tells a fight apart from a sweep. One name repeated is a duel, a long list in a few seconds is not.
Explosion TypeThe explosion involved, on every explosion check and on the two weapon checks that fire on an explosion.On AntiExplosionAi it is the value for WhitelistExplosions, which each of its Types entries carries its own copy of. On the weapon checks it is evidence, because those are keyed on the weapon rather than on the explosion.
MaximumThe ceiling the check was armed with, on the explosion scale and shake checks, beside the Scale or Shake value that crossed it.The pair says how far over the line the explosion was. The key is Types.Scale.Maximum or Types.Shake.Maximum, shipping at 1 and 3.
ValueThe voice range the client reported, on the voice spoofing check. The Type beside it is evidence rather than a Types entry.Compare it against that detection's own Maximum, which ships at 50. The panel does not carry the ceiling, so that one is read out of the config.
OldA property before the change Raven objected to, with New after it.Read this before assuming a cheat. A legitimate script changing a vehicle looks identical here.
Field names are the exact keys the resource stamps, so they are also what to search the log for.

A kick or ban an admin issued

These carry no detection context, so the panel records the action itself. Manual actions are also kept out of Raven's own detection feed on Discord.

FieldWhat it is
Issued ByThe dashboard user, the in-game admin, or the other resource that issued it.
Issued FromDashboard, In-game (the admin menu or rac in game), Server console, Another resource (one of its exports), or Game server when the game server did not say.
Admin IdentifierOn an in-game action, the identifier of the admin who issued it.

From a log entry to a config path

Config Option is the detection name as it appears in the config, and the Detections reference tells you which of the seven categories it sits in. Put those together and you have the path.

example
Config Option:    EntitiesSpawnProtection
Entity Type:      Objects
Entity Model:     -123456789
Entity Script:    ox_inventory

becomes

Detections.Entities.EntitiesSpawnProtection.WhitelistResources += "ox_inventory"

Note which lever that reaches for. Entity Model would also work as a whitelist entry, and it is the narrower of the two, but it only covers the one prop. Entity Script says every entity in the report came from your inventory resource, so trusting the resource fixes the whole class of report rather than this one model.

Working through one

Read Config Option first

It names the detection. If you do not recognise it, look it up in the Detections reference, which gives you the category and therefore the full config path.

Look for a cause you can name

Entity Script, Event Resource, Entity Model, Weapon Name. If one of these points at something of yours, that is your fix, and it is a whitelist entry rather than a threshold change.

Then read Type

If the detection has sub-checks and only one of them is wrong for your server, turning off that single Types entry keeps the rest of the detection working.

Check the evidence before you decide

If a screenshot or recording is attached, open it. For anything on the review list on Review a false ban, the evidence decides the case, not the detection name.

Make the change on the dashboard

Protection Rules for toggles, actions, thresholds and the whitelists, which are all inline on the detection's own card. Editing config.json in your resources folder does nothing, because the cloud sends the resource its config on every connect.

When a detection also files a network ban

A short list of detections escalates a system ban or kick to a network-wide ban on its own. If the detection you are reading is on that list, the player is also refused on other Raven servers that honour network bans. Lifting the ban on your dashboard takes the automatic network ban down with it, so clear a false positive promptly rather than leaving it. Global bans covers how escalation works.

When an event handler goes quiet instead of banning

Event protection can block an event without filing a detection, so there may be no log to find.

So the first symptom of an event false positive is usually a script that has quietly stopped working, not a ban. If a resource broke and there is nothing in the log yet, check whether its events are being blocked before you look anywhere else. Fixing it is covered on Server event protections.