Databases and files
Lists schemas, tables, views and columns in connected databases, and files in connected storage, through the account you create, read-only where the source allows it.
Rows 1 to 6 each name a database column or a file path.
Sensitive data discovery. Data classification.
The eight findings below come from one scan. Each names its source and location; a rejected candidate and two unread items are listed with the reason.
One scan of a seven-source estate: six databases and an HR document library. IRIS reads a sample of each table, and small tables in full, plus the files in scope; the last three rows show a rejection and two unread items.
| Source | Location | Detected type | Why | Sensitivity | Decision |
|---|---|---|---|---|---|
| CRM database | crm.person.email | Email address | Values match the email address pattern | Confidential | Approved by the reviewer |
| Core banking database | cbs.customer_master.pan | PAN (India) | Values have the PAN structure, including a valid holder-type letter | Restricted | Approved by the reviewer |
| HR database | hr.employee.ssn | US Social Security number | Nine digits in the SSN format | Restricted | Approved by the reviewer |
| Cards database | cards.card_customer.holder_name | Person name | Local name recognition on the sampled values | Confidential | Pending review |
| HR document library | Onboarding/scan_0412.pdf | Passport number | Text recognition read the scanned page, and the value has the passport number format | Restricted | Pending review |
| Orders database | shop.orders.order_number | Card number (candidate) | Values that pass the Luhn check, which random digits do about one time in ten, so a reviewer checked them | Low | Rejected by the reviewer as order numbers |
| HR document library | Payroll/salaries_2026.xlsx | Not classified | The file is password-protected, so IRIS couldn’t read it | Unknown | Not read: listed with the reason |
| Loans archive database | Whole source | Not classified | The connection failed, so no table was read | Unknown | Failed: recorded as a gap in the scan |
Dipal Panchal on skipping classification: “I will blindly give access to the entire finance database.”
Lists schemas, tables, views and columns in connected databases, and files in connected storage, through the account you create, read-only where the source allows it.
Rows 1 to 6 each name a database column or a file path.
Classifies values with pattern rules, checksums such as Luhn for payment cards and Verhoeff for Aadhaar, and local name recognition.
Rows 2 and 3 match a structure; row 4 uses name recognition; row 6 passed Luhn by chance.
Puts each finding in a review queue with a masked sample to check. The reviewer approves, reclassifies or declines it, and a decline needs a reason.
Rows 1 to 3 approved, rows 4 and 5 pending, row 6 declined with its reason.
Reads the text in scanned PDFs and images with text recognition, in English and Hindi, then classifies it like any other file. Results depend on scan quality.
Row 5 is a scanned passport page.
Records each table and file it couldn’t read, with the reason. An unread file shows as unknown, and isn’t counted as clean.
Rows 7 and 8: a password-protected file and a failed connection.
Adds custom patterns for identifiers specific to your organization, and can describe findings in your own sensitivity levels.
Not shown; you add patterns in the product.
Capabilities vary by edition.
Pick the sources and data types for this scan; anything left out goes on the record as a gap.
Give IRIS an account for each source that can only read, where the source allows it.
IRIS reads a sample of each table, and small tables in full, plus the files in scope.
Your reviewers decide each finding, then export it. Scans are point-in-time: schedule them, or run one on demand.
Findings are listed by database, schema, table and column, or by file. A reviewer or an auditor can go from any row to the place it came from.

The pattern library holds the built-in data types, grouped by category. Nine identifier types carry a checksum, such as payment cards, IBAN and Aadhaar, and IRIS validates it before it reports the type.

Last checked November 8, 2026, for IRIS v1. Coverage differs by source, so we confirm what IRIS can read in yours before the first scan.
103
built-in data types
For Indian, US and global formats; found by value or by name
9
identifier types validated by checksum
Cards (Luhn), IBAN, NPI, Aadhaar (Verhoeff); ask for the other five
12
database and warehouse engines
Including PostgreSQL, SQL Server, Oracle, MySQL and SAP HANA
15 min*
to scan a structured source
*About 100 tables, reading a sample of each table. Time depends on the number of tables and columns, not table size.
Findings by source and location, built from the connected systems themselves, with each one’s review status.
Each finding names the system it sits in and the data type it matched, in your own sensitivity levels if you set them. It covers connected sources only.
Each finding links to the scan that found it, and a re-scan adds a new dated result beside the earlier one. Unread sources and files are listed as gaps.
IRIS classifies the sources you connect. A system nobody connects stays outside the findings, so keep your list of unconnected systems beside the coverage record.
Each scan is a reading on its date, and IRIS doesn’t trace how data moved between systems. Your team fixes what a finding shows in the source system, and a re-scan records the new condition. An audit period needs repeated scans, and your auditor decides what the findings prove.
To follow one finding from the connected source to the evidence register, read the how-it-works page.
One scan followed from the connected source to the evidence register, step by step.
Turn findings into an evidence register and framework reports your auditor can trace.
The use case behind this scan: finding copies of customer data that a declared inventory missed.
No. IRIS reads a sample of each table, and small tables in full, so a value that sits in a few rows of a large table can be missed. The sampling method is recorded with each finding.
IRIS runs the Luhn check on each sampled value and counts only the values that pass. For random digits, about one in ten passes by chance, so the finding reaches a reviewer, who rejects it, as in the orders row above.
It depends on the column. Stored samples are masked by default for personal-data columns, so a reviewer checks those findings against a masked sample, and your administrator can mask every sample or switch storage off.
Yes. Text recognition reads scanned documents and images in English and Hindi, and files IRIS couldn’t read are listed with the reason.
Supported databases, cloud storage, Microsoft 365 and Google Workspace, and selected SaaS and code platforms. The integrations page lists each source and what IRIS can read there.
Pick the database your auditor always asks about. We’ll show you what IRIS finds there, and what it couldn’t read.