Unstructured Data

Unstructured data is information that has no predefined data model, so its meaning lives in the content itself. A customer table has columns that say what each value means. A signed contract, a chat thread or a recorded call does not come with such labels.

Most business information is of this kind. Because it can hold client records, price lists or personal details, Data Loss Prevention (DLP) programs have to inspect what a file says, not only where it is stored.

Data falls into three groups based on how much structure it carries. Unstructured data includes Word and PDF files, email bodies, chat logs, presentations, images, scanned contracts, design files, and audio and video recordings.

It is everywhere in a typical organization. Industry analyses consistently estimate that 80–90% of new corporate data is unstructured, and that share keeps growing as collaboration moves to documents, chats, and media.

Messaging is one of the largest sources. Much of it moves through electronic messaging services, where a single thread can hold files, screenshots, and pasted text together.

• Shared drives and document repositories
• Email mailboxes and messenger histories
• Cloud file-sharing services
• Scanned documents attached to business processes
• Photos of whiteboards and screens

Structured data can be protected at the source. A database has a boundary, permissions, and audit logs. Unstructured data has none of that, and most accidental disclosures and insider data theft involve it moving through everyday channels.

Four traits explain why: it has no schema or labels, it copies freely, it comes in many formats, and it belongs to everyone and no one, so central catalogs rarely cover it.

Because unstructured data cannot be controlled at a single source, controls focus on content analysis where the data moves.

Text is classified with keyword dictionaries and thesauri, such as a dirty word list. Hash values and digital fingerprints match exact documents and their copies, and OCR extracts text from scans and images so they can be analyzed like any other file.

Labels then let downstream policies enforce rules on email, web, messengers, and removable media.

The console keeps these methods in one place. The system thesauruses hold 35 keyword sets, such as SSN, Credit Card, and Financial, and each one shows its word count and search hits. Image recognition decides which channels and file formats are read as documents, and OCR covers scans.

These settings belong to the Anexet DLP module, which applies them to documents, messages, and files in motion.

1 / 2

• 
• 

Start with the content that would hurt most if it left the company, then trace the channels it actually uses. Rules written for every file at once tend to block ordinary work and get switched off.

Measure how much unstructured data exists first, then decide where it has to be read: where it is created, shared or sent out.