ibm_verify_events_raw.adaptive_access
Every IBM Verify adaptive access evaluation: the risk level IBM Verify assigned to an access attempt, and what its policy did about it. Prefer the daily summary, events_adaptive_risk_summary.
Keywords: adaptive access, risk, risk level, policy, device, browser, location, bots, raw events
Provenance
Pulled hourly from IBM Verify by the streaming job. See IBM Verify Events for what is hashed and how to filter by day.
Data starts on 2026-09-09, when adaptive access went live.
- Updated: Hourly
- Retention: 3 years
- Steward: EDCP Data and Research Team
- Contact: Yinlin Zhang
- Location:
s3://<raw-bucket>/ibm_verify_events_raw/adaptive_access/
Fields
One row per evaluation.
| Field Name | Type | Description |
|---|---|---|
| time | string | When the evaluation happened, in UTC (for example, “2026-09-15 00:00:05 UTC”) |
| userid | string | Hashed user ID |
| username | string | Hashed username |
| applicationid | string | IBM Verify’s numeric ID for the application being reached |
| applicationname | string | Application name (for example, “IRIS - PRC - Prod”) |
| risk_level | string | “LOW”, “MEDIUM”, “HIGH”, “VERY_HIGH”, or empty |
| decision_reason | string | IBM Verify’s reason for the rating (for example, “Access from a known and trusted device”). Empty when there is no risk level |
| policy_action | string | What the policy did. Mostly “ACTION_CONTINUE”, “ACTION_MFA_PER_SESSION”, “ACTION_MFA_ALWAYS”, or “ACTION_REDIRECT” |
| policy_id | string | IBM Verify’s ID for the access policy |
| policy_name | string | Access policy name (for example, “GCS_AccessPolicies_Default_Migration”) |
| browser | string | Browser (for example, “Firefox”) |
| os | string | Operating system (for example, “Windows NT”) |
| city | string | City from the IP address |
| region | string | Province, territory, or state from the IP address |
| country | string | Country from the IP address, as ISO 3166-1 alpha-3 (for example, “CAN”) |
| realm | string | IBM Verify directory, “cloudIdentityRealm” |
| origin | string | Hashed client IP address |
| year | string | Partition key. UTC year, for example “2026” |
| month | string | Partition key. UTC month, for example “09” |
| day | string | Partition key. UTC day of the month, for example “15” |
Notes
Some evaluations have no risk level. Almost all of them have policy_action = "ACTION_REDIRECT". The summary table counts them in total_invocations but in no level.
Location columns are named differently here. They are city, region, and country, not the geoip__ columns of the other three tables.
Example. Evaluations and distinct users by risk level on an ET day:
tbl(con, in_schema("ibm_verify_events_raw", "adaptive_access")) |>
# partitions are UTC days, so read the next one too for the ET evening
filter(year == "2026", month == "09", day %in% c("15", "16")) |>
mutate(et_day = sql("date(with_timezone(CAST(substr(time, 1, 19) AS timestamp), 'UTC') AT TIME ZONE 'America/Toronto')")) |>
filter(
et_day == as.Date("2026-09-15"),
!is.na(userid), !userid %in% c("", "UNKNOWN")
) |>
group_by(risk_level) |>
summarise(evaluations = n(), unique_users = n_distinct(userid))