Definition
Re-Authentication Proportionality Index (RAPI) measures the correlation between re-authentication trigger events and actual risk level via post-hoc assessment of a sample.
In plain language: take a sample of re-authentication events.
The Measurement Gap
A large enterprise rolled out AI-driven adaptive authentication 14 months ago. The platform evaluates each session continuously, scoring risk based on device, location, behavior, time, and other signals, and triggering re-authentication when the score exceeds a threshold. The platform marketing materials describe it as "frictionless when safe, protective when risky." The deployment metrics report re-authentication rates that the security team considers appropriate.
Then the IT help desk surfaces a pattern. Help desk tickets related to authentication friction are up 80% year-over-year. Many are from users in roles that perform routine work (sales, support, operations) who report being re-authenticated multiple times per day for actions that are part of their normal workflow. Some users have started sharing sessions (one person logs in and others use the open browser) to avoid the re-auth friction. The security team had not been tracking the friction; the platform's risk scores were the operational metric.
A security analyst investigates a sample of re-auth events. Of 50 random re-auths from the past week, careful post-hoc assessment finds that 12 corresponded to genuinely elevated risk (unusual location, device change, time-of-day anomaly), while 38 corresponded to normal user behavior that the AI risk scorer had flagged for reasons that did not, in retrospect, indicate elevated risk. The platform was producing security friction that did not protect against actual risk, and the friction was driving users to insecure workarounds.
The legacy identity metrics report a healthy program.
Re-authentication rate: tracked, considered appropriate.
MFA fatigue alerts: not tracked.
Session sharing detection: minimal coverage.
Authentication success rate: stable.
Compromise prevention rate: tracked through other measures, considered adequate.
None of these report on whether the re-authentication triggers are proportionate to actual risk. The re-auth rate measures how often re-auths happen. The fact that most re-auths are not proportionate to risk is not in the metrics; it requires separate investigation that the team had not been doing.
These metrics were designed for static authentication policies where re-auth was triggered by specific known events (session timeout, sensitive action). With AI-driven adaptive authentication, the triggers are dynamic and based on risk scoring that may or may not be calibrated.
There is no metric that tells you whether AI-driven re-authentication decisions are proportionate to actual risk or are driving users toward insecure workarounds, which means security enforcement looks strong while credential hygiene degrades.
Formula
Re-Authentication Proportionality Index (RAPI) measures the correlation between re-authentication trigger events and actual risk level via post-hoc assessment of a sample. In plain language: take a sample of re-authentication events. For each, conduct post-hoc assessment to determine whether the session was genuinely at elevated risk or whether the re-auth was friction without security benefit. RAPI is one minus the fraction of re-auths assessed as low-risk. A value near 1 means most re-auths protected against genuine risk. A value near 0 means most re-auths were unnecessary friction.
The formula:
RAPI = 1 − (Low_risk_reauths / Total_reauths)
Where:
- Low_risk_reauths = re-authentication events where post-hoc assessment determined the session was genuinely low-risk
-
Total_reauths = all forced re-authentication events in the sampleUnits are a value from 0 to 1. 1.0 means every re-auth was proportionate. 0 means all re-auths were on low-risk sessions. A worked example. The enterprise runs RAPI on a sample of 200 re-auth events from the past month:Post-hoc assessment categorizes each event as genuinely-elevated-risk or low-risk-friction:
-
67 events: genuinely elevated risk (unusual location, device change, anomalous time, suspicious behavior)
- 133 events: low-risk (routine activity within normal patterns, no observable risk indicator)
RAPI = 1 − (133 / 200) = 1 − 0.665 = 0.335
Per the catalog guidance (above 0.60 mostly genuine risk, below 0.30 mostly friction) this RAPI of 0.335 sits in the warning band. The action trigger at 0.40 has been crossed. The team needs to raise the risk threshold for re-auth triggers and supplement the AI risk scoring with rules that explicitly require corroborating evidence before triggering re-auth. Three edge cases need explicit treatment. First, post-hoc risk assessment has its own uncertainty: the assessor is making a judgment about whether the session was at risk based on available evidence. Use multiple assessors and a structured assessment framework to reduce subjectivity. Second, measure workaround behavior alongside RAPI: session sharing, browser extension password managers used in non-standard ways, increased password reset frequency. The workaround indicators surface security cost that low RAPI predicts. Third, the minimum sample is 200 re-auth events per measurement period for stable percentages. Cross-references. Upstream of RAPI is Behavioral Baseline Integrity Score (BBIS). If the behavioral baseline is contaminated or drifting, the risk scores built on it will be miscalibrated, and RAPI will surface the operational consequence. Downstream is
Practitioner Effectiveness Uplift (PEU, GMI-IT-030): excessive re-auth friction reduces practitioner effectiveness; RAPI is one input to understanding why PEU may be degraded. The misinterpretation to watch for: a team reports RAPI = 0.7 (above the threshold) and concludes the re-auth triggers are proportionate. They miss that the sample was drawn primarily from re-auths in high-risk contexts (privileged account access, sensitive data access), where most re-auths are correctly proportionate. The routine user re-auth sample, which is much larger in volume, has RAPI of 0.3. Sample re-auths by user-context type and report RAPI per context.
Proposed formula, pending validation pilots.