The report covers declared systems
In Dipal Panchal’s talks with chief information security officers (CISOs), the systems missing from their reports are test and development environments, small applications and tools bought on a card.
Sensitive data discovery. Shadow data.
A restored staging database holds production’s customers and is often missing from the inventory. IRIS finds such copies in the sources you connect.
In Dipal Panchal’s talks with chief information security officers (CISOs), the systems missing from their reports are test and development environments, small applications and tools bought on a card.
Consent is stored in the application that collected the data. Of carrying it into replicas and the warehouse along with the data, Dipal Panchal says: “Ideally, no company does this.”
Test and development databases are generally left out of the compliance program, as Dipal Panchal puts it. A restore from a production backup brings real customer records into that test database.
Records from one scan across six places copies collect. IRIS finds the sensitive columns and files; your team adds the reason each is a copy, and the owner’s decision.
| Source | Location | Detected type | Why | Sensitivity | Decision |
|---|---|---|---|---|---|
| Staging database | crm_staging.person.email | Email address | Restored from a production backup for testing; the rows match crm.person | Confidential | Owner: CRM lead. Replace with test values, then re-scan |
| Analytics warehouse | analytics.customer_snapshot.aadhaar | Aadhaar number | Loaded nightly from the core banking database, without its consent fields | Restricted | Owner: data platform lead. Keep; limit reads to two named roles |
| Amazon S3 bucket | ops-exports/2026-10/cards_reconcile.csv | Payment card number | A support export for a reconciliation whose ticket closed in October | Restricted | Owner: payments operations. Delete the file and re-scan the bucket |
| SharePoint site | Finance/Shared/loan_extract_q3.xlsx | Bank account number | Copied out of the loans system for a quarter-end review | Restricted | Owner: finance controller. Move it back into the loans system |
| QA database | qa_checkout.payments.card_number | Payment card number | The values are processor test numbers such as 4111 1111 1111 1111, not customer cards | None | Rejected by the reviewer, with that reason |
| Laptops | Downloads and Desktop folders | Not scanned | IRIS doesn’t scan endpoints; files in synced OneDrive folders are scanned instead | Unknown | Excluded: recorded as a gap for the endpoint team |
Tables and files IRIS couldn’t read are listed as not read, with the reason, which keeps a locked spreadsheet from passing as clean.
Once the owner deletes or moves a copy, a re-scan of the same source is a new scan with its own date, and the scan history keeps the earlier one.
Point it at the places copies collect, such as staging databases, export buckets and SharePoint sites, and at the system of record.
Covers 12 database and warehouse engines, including PostgreSQL, SQL Server, Oracle, MySQL, Snowflake, Databricks, BigQuery and SAP HANA. IRIS reads a sample of each table, and small tables in full.
Reads PDF, current Word, Excel and PowerPoint, HTML, text, CSV and JSON files in connected cloud storage, Microsoft 365 and Google Workspace.
Validates nine identifier types by checksum before reporting them, such as Luhn for payment cards and Verhoeff for Aadhaar. Its 103 built-in data types cover Indian, US and global formats.
Text recognition reads scanned PDFs and images in English and Hindi, on the IRIS host. A file it can’t read gets a reason and is never counted as clean.
Capabilities vary by edition.
IRIS finds sensitive data where it sits. It doesn’t trace lineage, so it can’t say which table a copy was loaded from; the owner records that, as in the “why” column above.
It sees only the sources you connect, and it doesn’t scan laptops. Because it reads a sample of each large table, a value held in a handful of rows can be missed in any one scan.
GDPR Article 30: personal data categories. HIPAA §164.308(a)(1)(ii)(A): health data locations. PCI DSS 12.5.2: account data locations, at least every 12 months.
For UK GDPR, UAE PDPL, Saudi PDPL and PCI DSS, IRIS maps discovery findings to their controls. A scored readiness view for discovery work isn’t in v1; ask us if you need one.
Scan the bucket your exports land in, then the database they come from, and compare the two scans’ findings: each customer table found in both is a copy to add to the inventory. The how-it-works page follows one scan from connection to findings.
No. IRIS finds the sensitive data in each source, but it doesn’t trace lineage. Your team records where a copy came from; for a database column, that goes in the review reason.
No. IRIS scans only the sources you connect. List the others as gaps beside the coverage record, so the scan’s limits stay on the record.
Findings, coverage records and stored samples, on your host unless you turn on emailed reports. It reads values in memory. Stored samples are masked by default in personal-data columns. Explore architecture
Restricted data that isn’t encrypted and has no access policy, in Dipal Panchal’s order. In his words, “that is the biggest risk you have to solve right now.”
Name the database your exports are cut from. We’ll show what IRIS finds in it, and in a demo, the copies around it.