§ 05 · PIP — The identity fabric
Powered by Radiant Logic
A single, current Golden Record per person.
Without a complete profile, the system cannot make good decisions. The PIP ensures every decision uses the full picture — never a stale or incomplete record.
i · Identity
Who they are
Name, role, department, clearance level, hire date.
ii · Organization
Who they answer to
Manager, cost center, division.
iii · Location
Where they work
Office, city, country, timezone, IP range.
iv · Device
What they're using
Type, encryption status, managed or personal, last seen.
v · Risk
How risky they are
Current risk score, peer group, specific risk factors.
vi · Access
What they can do
Every active permission, source application, risk level.
How peer groups are discovered
Radiant Logic FPMax algorithm — four steps from raw employee data to named peer groups.
- Data prep Polars conversion, drop constant/all-unique columns → 64 employees × ~20 attribute & permission columns
- Itemset mining FPMax algorithm, min_support = 0.05, gamma = 0.5, weighted features (entitlements 2.0×, structural 0.5×) → Ranked maximal frequent itemsets
- Hierarchical clustering Linkage-based merge at similarity ≥ 0.4 → 21 natural peer groups
- Outlier scoring Composite penalty: unassigned (+3.0), multi-cluster (+1.0/extra), low-support (+1.0 to +2.0) → 25 of 64 flagged