Open source intelligence

What the term covers

OSINT stands for open source intelligence. It means collecting, cross-checking and interpreting information you can reach without circumventing anything: a company register, a press article, a legal notice, a public page, a land database, a public-domain satellite image.

The word “open” refers to access, not to price nor to ease. A paid database anyone can subscribe to is still an open source. Information you would have to obtain by impersonating someone, or by entering a system without authorisation, is not one, even if it is lying around online.

The word “intelligence” matters just as much. Collection is not yet intelligence. What makes the difference is the cross-checking, the connecting and the interpretation that come afterwards.

What it is not

Three confusions come up regularly.

Intrusion. Guessing a password, exploiting a flaw, logging into an account that is not yours: that is not OSINT, it is an offence. The boundary is not blurry, it runs along authorisation of access.

Social engineering. Posing as a candidate, a customer or a journalist to obtain information belongs to another discipline, with another framework. A source stays open as long as you did not have to lie to reach it.

Automated scraping. Hoovering up a site and piling up thousands of rows produces volume, not understanding. Without cross-checking or ranking, all that remains is a table nobody will read.

Where the information comes from

The families of sources vary by country, but the structure holds.

• Official registers. Company registers, legal notices, official gazettes, land data, public procurement. These are the strongest sources, because they are dated and enforceable.
• Press and publications. Articles, statements, reports from organisations, academic work. Useful for context and for dating a fact.
• Technical traces. Domain registrations, certificates, server addresses, page archives. They connect objects when documents do not.
• What people and organisations publish. Websites, professional profiles, job adverts, public appearances. Handle with care: these are claims, not established facts.
• Imagery and mapping. Satellite views, street photography, topographic surveys, within the limits of their terms of use.

How the work proceeds

The work almost always comes down to the same loop. You start from something known, you look for what it attaches to, you verify, then you set off again from the new item. A pivot is the piece of information that allows that jump: an email address present on two sites, a postal address shared by two companies, a handle reused from one platform to another.

The difficulty is not finding, it is keeping. After a few days a case holds dozens of items, and nobody remembers offhand why a given company was attached to a given person. That is the role of investigation mapping: fixing the objects, naming the links and keeping the reasoning readable. The four-step method describes that loop in detail.

What makes the work sound

Four criteria let you judge an open source case, yours or someone else’s.

1. Every claim has an identifiable source. Not “it is on the internet”, but a document, a date, a reference.
2. Facts are separated from hypotheses. A likely link announced as likely discredits nobody. A likely link announced as certain brings down everything else with it.
3. Cross-checks are independent. Two sites copying each other do not make two sources.
4. The reasoning can be replayed. Someone else, with the same file, must be able to walk the same path and arrive at the same place.

The limits

An open source is dated, sometimes out of date, often incomplete. A register reflects a declaration, not necessarily reality. An absence of information is not proof of absence.

Then there is the legal framework. Public information remains personal data as soon as it concerns an identifiable person, with the obligations that follow. The guide on personal data sets out what that means in practice.

Finally there is the question of volume. Nothing obliges you to collect everything you can reach. A tight file that answers the question asked will hold up better than an exhaustive one nobody masters.