In our earlier Dynamic AI Safety weblog and the underlying CAMLIS 2025 paper, we described a launch platform constructed to maneuver new protections into manufacturing with out disrupting buyer workflows. Such a platform is a requirement for safety techniques as a result of the fixed evolution and adaptation of adversaries require the same response loop from distributors like Cisco.
Speedy detection-model churn creates potential downstream disruptions for purchasers, who all of the sudden, and with out understanding it, begin utilizing a more moderen model of the mannequin that behaves in a different way from the earlier iteration. One main variable between deployments is mannequin aggression: the newer mannequin is healthier however can also be deployed extra aggressively, breaking workflows that appeared tremendous only some hours in the past.
Preserving the aggression stage between releases is one thing we concentrate on throughout each mannequin replace. If a buyer chooses a blocking tier that flags about 1 in 1,000 requests, a mannequin retrain shouldn’t silently flip that into 1 in 200 or 1 in 20,000. The detector might enhance beneath, however the buyer’s false-positive funds ought to stay the identical.
Completely different downstream customers additionally function at totally different factors on that tradeoff. A SOC operating aggressive blocking sits removed from a software that solely enriches logs, so it’s not sufficient to protect one threshold throughout releases. The entire vary of working factors, from very aggressive to very conservative, has to hold the identical which means from one mannequin model to the following.
It is a widespread downside for anybody deploying detection fashions, so we’re open-sourcing our resolution for FPR calibration that may be utilized earlier than a mannequin is launched to reduce the prospect of buyer disruption. The strategy works offline on benign scores and ships a bounded sklearn artifact together with the mannequin. The code is at github.com/cisco-ai-defense/fpr-model-calibration, and the paper describing the technical particulars will be downloaded right here.
Why false-positive fee is the proper contract
False-positive fee (FPR) is the fraction of benign site visitors that might be flagged at a threshold. For a mannequin rating threshold, FPR estimates how a lot authentic exercise the brink will interrupt in manufacturing.
FPR calibration differs from chance calibration, which estimates Pr(assault | rating). For a lot of safety fashions, that chance depends upon an assault distribution that’s uncommon, adversarial, and quickly shifting. Attackers change techniques when detectors enhance. The constructive class a mannequin sees throughout coaching is due to this fact a report of previous assaults, not a secure pattern of future assaults.
FPR calibration relies upon solely on benign site visitors. In lots of manufacturing safety settings, benign site visitors is extra plentiful, simpler to measure, and tied on to false-positive hurt. If the calibrated rating says a request is a 1-in-1,000 benign occasion, the product crew can purpose about alert quantity without having to know tomorrow’s assault prevalence.
What a calibrated rating means
The calibrator maps uncooked mannequin scores onto a hard and fast rating contract. The calibrated rating contract maps widespread working tiers to focus on FPRs:
The dimensions is logarithmic as a result of manufacturing FPR selections are logarithmic. Transferring from 1% to 0.1% and from 0.1% to 0.01% are each tenfold reductions in benign alerts. A linear rating axis would compress the low-FPR area coated by the 0.50, 0.70, and 0.85 working tiers.
With the rating contract in place, coverage thresholds keep secure throughout mannequin releases. A coverage can block at 0.50, alert at 0.30, and enrich logs at 0.10. When the mannequin crew ships a brand new detector model, it ships a brand new calibrator with it. The coverage thresholds hold their FPR which means regardless that the uncooked mannequin scores beneath modified.
How a lot information is sufficient?
One widespread gotcha when estimating the efficiency of detection fashions is simply how a lot information you really must correctly calibrate, and even measure, a mannequin. Whereas assaults can appear to be all over the place in public check units, in follow they’re very uncommon, normally under 0.1% of site visitors. At these charges, the mannequin must be extraordinarily correct to maintain the false-positive fee sensible, and calibrating it requires much more benign information than one would count on.
A traditional-approximation rule of thumb offers about 16 / p benign samples for plus-or-minus 50% relative precision at 95% confidence, the place p is the goal FPR. For widespread working factors, the tough pattern counts are:
Pattern measurement dominates low-FPR error in follow, and extra benign information is the one path to tighter estimates.
Validation on a public benchmark
We validated the tactic on the general public Credit score Card Fraud Detection benchmark (284,807 transactions, 492 fraud instances), becoming the calibrator on a held-out benign subset:
The takeaway is easy: so long as the benign distribution stays pretty fixed between calibration and manufacturing, a mannequin will be calibrated very precisely.
What modifications for product groups
An FPR-calibrated launch contains the detector, the calibrator, and both calibrated-score serving or uncooked thresholds derived from the calibrator. Coverage thresholds hold their FPR which means, clients hold their false-positive funds, and the mannequin can enhance beneath.
The identical contract additionally makes detector scores simpler to check throughout classes. If a prompt-injection detector and a data-leakage detector each emit calibrated rating 0.50, every rating means the identical factor about benign rarity. Compound insurance policies nonetheless want their very own FPR measurement, however their inputs not combine unrelated uncooked rating scales.
Getting began
Match the calibrator with fit_calibration_pipeline:
from fpr_model_calibration import fit_calibration_pipeline
import joblib
pipeline = fit_calibration_pipeline(benign_scores, n_knots=10000)
joblib.dump(pipeline, “calibration.pkl”)
Manufacturing inference calls the serialized sklearn pipeline:
pipeline = joblib.load(“calibration.pkl”)
calibrated = pipeline.predict(raw_scores.reshape(-1, 1))
FPR calibration offers mannequin releases a secure rating contract with out changing contemporary benign information, drift monitoring, or detection-quality analysis. For safety techniques that retrain underneath adversarial stress, that contract lets detectors enhance whereas coverage thresholds hold their FPR which means.
Hyperlink to the open supply GitHub repo will be discovered right here:
https://github.com/cisco-ai-defense/fpr-model-calibration
and the preprint:
https://arxiv.org/abs/2607.05481
