Risk Scoring

Every identity in VeraID carries a dynamic risk score ranging from 0 to 100. This score is continuously recalculated based on behavioral signals, credential hygiene, access patterns, and policy compliance. Risk scores drive automated responses such as step-up verification, access restrictions, suspension, and alerting.

How Risk Scoring Works

VeraID evaluates each identity against a set of weighted risk factors. When a factor’s conditions are met, its weight is added to the identity’s score. The final score is capped at 100 and rounded to the nearest integer.

Risk scores are recalculated:

• After every authenticated action
• On a scheduled basis (every 15 minutes for active identities)
• When credential, policy, or ownership changes occur
• When anomaly detection triggers fire

Risk Levels

Scores map to four risk levels that determine the severity of automated responses and dashboard indicators.

Score Range	Level	Dashboard Indicator	Default Behavior
0 – 24	Low	Green	Normal operations
25 – 49	Medium	Yellow	Flagged for review, optional notifications
50 – 74	High	Orange	Step-up verification, access restrictions triggered
75 – 100	Critical	Red	Automatic suspension, incident alert, SOC notification

Tip

Risk level thresholds and automated responses are fully configurable via Policies. You can set custom thresholds per identity type, group, or environment.

Standard Risk Factors

The following nine factors apply to all identity types. Each factor has a weight (the number of points added to the score) and a trigger condition.

Failed Authentication Attempts

Weight	+25
Trigger	10 or more failed authentication attempts within a 30-day window
Rationale	Repeated failures may indicate a credential stuffing attack, brute-force attempt, or misconfigured client

Policy Violations

Weight	+25
Trigger	10 or more denied policy evaluations within 7 days
Rationale	Frequent policy denials suggest the identity is attempting unauthorized actions or has drifted from its intended scope

Geographic Anomaly

Weight	+20
Trigger	Authentication from an IP address in a geographic region not previously associated with this identity
Rationale	Unexpected geographic shifts may indicate credential theft or proxy-based access from an unauthorized location

Dormant Reactivation

Weight	+20
Trigger	Identity inactive for 30 or more days, followed by 10 or more actions within 24 hours
Rationale	A sudden burst of activity from a long-dormant identity is a strong indicator of compromise

Unusual Request Rate

Weight	+20
Trigger	Request rate exceeds 3x the identity’s rolling 30-day average
Rationale	Traffic spikes well above baseline may indicate automated abuse, data exfiltration, or a compromised credential

Credential Not Rotated

Weight	+15
Trigger	Any associated credential has not been rotated in 90 or more days
Rationale	Long-lived credentials increase the window of exposure if compromised; regular rotation is an industry best practice

Unusual Access Hours

Weight	+15
Trigger	50% or more of requests occur during off-hours (10:00 PM – 6:00 AM in the identity’s configured timezone)
Rationale	While some automation legitimately runs overnight, a shift toward off-hours access can indicate unauthorized use

Expiring Credential

Weight	+10
Trigger	Any associated credential expires within 7 days
Rationale	Impending expiration increases operational risk; teams should be alerted to rotate before an outage occurs

Excessive Permissions

Weight	+10
Trigger	Identity has more policy bindings than the 90th percentile for its type
Rationale	Over-permissioned identities expand the blast radius if compromised; least-privilege principles should be enforced

AI Agent-Specific Risk Factors

Identities of type AI_AGENT are evaluated against five additional risk factors that address the unique threat surface of autonomous AI systems.

Caution

AI agent risk factors are cumulative with standard factors. An AI agent that triggers prompt injection detection (+30) and has unrotated credentials (+15) would receive a combined score of 45, placing it in the Medium risk level.

Prompt Injection Detected

Weight	+30
Trigger	VeraID’s prompt injection detection system identifies a potential injection attempt in the agent’s input or output
Rationale	Prompt injection is a critical threat to AI agents; a detected attempt may indicate an active attack on the agent’s decision-making process

Capability Abuse

Weight	+25
Trigger	5 or more denied capability requests (MCP tool calls, API actions outside allowed scope)
Rationale	Repeated attempts to use unauthorized capabilities suggest the agent has been manipulated or is behaving outside its intended parameters

Budget Exceeded

Weight	+20
Trigger	Agent has consumed more than 80% of its configured budget for the current period
Rationale	Runaway spending may indicate an infinite loop, prompt injection causing excessive API calls, or unauthorized task escalation

Unusual Model Usage

Weight	+15
Trigger	Agent has used 3 or more different AI models within a single evaluation period
Rationale	Unexpected model switching may indicate an attacker probing for vulnerabilities across model providers or attempting to bypass model-specific safety controls

High Cost Rate

Weight	+15
Trigger	Current daily spending rate exceeds 2x the identity’s rolling daily average
Rationale	A sudden increase in cost rate — independent of total budget consumption — signals anomalous behavior that warrants investigation

Risk Factor Summary

The table below provides a complete reference of all risk factors sorted by weight.

Factor	Weight	Category	Trigger Summary
Prompt Injection Detected	+30	AI Agent	Injection attempt detected
Failed Authentication	+25	Standard	10+ failures in 30 days
Policy Violations	+25	Standard	10+ denied evaluations in 7 days
Capability Abuse	+25	AI Agent	5+ denied capability requests
Geographic Anomaly	+20	Standard	Request from new geographic region
Dormant Reactivation	+20	Standard	30+ days inactive, then 10+ actions in 24h
Unusual Request Rate	+20	Standard	3x above 30-day average
Budget Exceeded	+20	AI Agent	>80% of budget consumed
Credential Not Rotated	+15	Standard	90+ days without rotation
Unusual Access Hours	+15	Standard	50%+ requests during 10 PM – 6 AM
Unusual Model Usage	+15	AI Agent	3+ models used in one period
High Cost Rate	+15	AI Agent	2x above daily spending average
Expiring Credential	+10	Standard	Credential expires within 7 days
Excessive Permissions	+10	Standard	Permissions above 90th percentile

Viewing Risk Scores

Dashboard

The VeraID dashboard displays risk scores with color-coded indicators on the identity list view, identity detail page, and the risk analytics overview. Identities in the High or Critical range are surfaced in the priority queue on the dashboard home page.

API

Retrieve an identity’s current risk score as part of the identity object:

curl -X GET https://app.veraid.io/api/v1/identities/idt_3a1f8c29-b7d4-4e2a-9c8f-1d5e7a2b4c6d \
  -H "Authorization: Bearer kd_live_your_api_key"

The riskScore field in the response contains the current score:

{
  "id": "idt_3a1f8c29-b7d4-4e2a-9c8f-1d5e7a2b4c6d",
  "name": "payment-processor",
  "riskScore": 35,
  "status": "ACTIVE"
}

Configuring Risk-Based Policies

Risk scores integrate directly with the VeraID policy engine. You can create policies that trigger automated actions based on risk thresholds:

{
  "name": "auto-suspend-critical-risk",
  "type": "RISK_THRESHOLD",
  "conditions": {
    "riskScore": { "gte": 75 },
    "identityType": ["SERVICE_ACCOUNT", "API_KEY", "CI_CD"]
  },
  "actions": [
    { "type": "SUSPEND_IDENTITY" },
    { "type": "NOTIFY", "channels": ["email", "webhook"] },
    { "type": "CREATE_INCIDENT", "severity": "critical" }
  ]
}

Note

Risk-based policies are evaluated in real time whenever a risk score is recalculated. There is no delay between a score change and policy enforcement.

What’s Next

• Identity Groups — Apply risk-based policies to groups of identities
• Policies Overview — Build policies that respond to risk scores and behavioral signals
• Audit Logs — Review the audit trail for risk score changes and automated actions