Published on · 5 min read
What a hard drive reveals — and what it never proves
A workstation keeps a record of what was installed, run, plugged in and connected. Here is what an analysis can actually establish, what it will never be able to say, and the reflexes that destroy the evidence before anyone has started.
A company director rarely calls me out of curiosity. They call because a workstation is behaving strangely, because an employee left with unanswered questions, or because a doubt has settled in and they do not know how to resolve it without making things worse.
The good news is that a computer forgets very little. The bad news is that the first few minutes after the doubt arises usually decide everything.
What a disk keeps
A Windows system keeps a far more detailed record of activity than most users realise.
Installed software leaves several independent traces: an uninstall entry in the registry, a fingerprint in the compatibility database, an installation folder. Even once removed, software usually leaves a scar.
Programs that actually ran are recorded elsewhere, in a start-up optimisation mechanism. It holds the date of first execution, the date of the last one, and the number of launches. That is a different fact from installation, and the distinction matters: an executable launched from a USB stick was never “installed”.
Logon sessions are journalled along with their type. A session opened at the keyboard and one opened remotely do not carry the same code. That is precisely what makes it possible to say that a machine was used remotely, and when.
Remote access tools keep their own connection logs, including identifiers of the machines that connected.
USB devices leave their make, model and serial number, with the dates of first and last connection.
Network usage per application is retained for several weeks in a little-known system database. It shows which application transferred data, and at what time.
What a disk will never say
This is the part that serious reports state plainly, and that others quietly skip.
It will not say who was at the keyboard. A session opened on an account proves that an account was used. If the password was written in a notebook, shared between two people or saved in the browser, the account identifies no one. Bridging the gap between an account and a person is a legal assessment, not a technical finding.
Absence of evidence is not evidence of absence. Event logs are circular: past a certain volume, older entries are overwritten. A clean-up, a reinstall or simply the passage of time is enough to erase what you are looking for.
Timestamps deserve caution. One log may record in universal time while another records local time. Two hours misread are enough to place someone in front of a screen when they were somewhere else. A report must state the time zone used and the conversion applied — it is the first thing an opposing expert will attack.
The reflexes that destroy evidence
They are almost always well-intentioned, and committed within the hour.
Switching the machine back on “just to look”. A Windows boot writes hundreds of files, updates access times and rotates logs. What you were looking for can disappear while you look for it.
Copying the interesting files onto a USB stick. An ordinary copy loses the original dates and proves nothing about where it came from.
Handing the machine to the nephew who knows about computers. His goodwill is real; his method will not stand up.
The right reflex is the opposite of instinct: do nothing. Unplug it, put it away, write down the date and time, and make the call.
A method that holds
There is nothing spectacular about it, and that is exactly the point.
The equipment is photographed, then placed under a numbered seal. The storage device is copied identically, sector by sector, without ever being booted. Cryptographic hashes of the device are taken before and after the work: if they match, it is demonstrated that nothing was altered. The analysis is then carried out exclusively on the copy, and every operation is recorded in a timestamped log that accompanies the report.
The original device goes back under seal and is returned to you against a signed receipt.
The legal framework is not optional
An employer may verify how their own equipment is used. That power stops in three specific places.
The measure must be proportionate to the aim pursued. Files an employee has identified as personal cannot be opened outside their presence, except in particular circumstances — that is settled case law in France. And if employees were never informed that monitoring exists, the evidential value of what you find is weakened, even though the technical analysis remains possible.
Serious work therefore begins with a written mandate signed by the owner of the equipment, and with three simple questions: is there an IT policy, have employees been informed, and has the works council been consulted on monitoring measures?
Finally, if proceedings are seriously contemplated, a formal record drawn up by a commissaire de justice at the time the machine is opened carries evidential weight that a private report, however rigorous, never will. Saying so beforehand beats learning it afterwards.
In short
A hard drive is a precise but literal witness: it describes what a machine did, not what a person intended. A good forensic report is recognisable by the fact that it owns that limit — it describes timestamped facts, cites its technical sources, states what it could not establish, and leaves the legal assessment to those whose job that is.
It is less compelling than a clear-cut conclusion. It is far more solid on the day someone challenges it.