1. What exfiltration means and why it goes unnoticed
Data exfiltration means the unauthorised transfer of data out of your systems. It can happen through a compromised account, an infected device, a supplier with excessive access, or an employee copying information onto a personal medium. Unlike ransomware, it does not block anything: it often leaves no visible sign, and the theft may only come to light through an external report.
The first problem is establishing that a theft has actually taken place, because it often leaves no obvious trace. The second is understanding which data has left, whose it is, and whether the leak is relevant under the law. The playbook must answer both questions without assuming the event has been contained. Even news from a customer or an outside researcher must be treated as a signal to verify.
2. Prerequisites: knowing what you hold and where
You cannot assess a theft you cannot describe. Before any incident you need a record of processing activities: which categories of information you hold, in which systems, who accesses them and for what purpose. The record required by the GDPR, the European data protection regulation, is a good starting point, because it forces you to map this out. It needs to be kept up to date as things change, not prepared once and forgotten.
You also need access logs kept long enough to allow analysis. If logs are deleted after a short time, or do not exist for the most sensitive systems, the impact assessment becomes an estimate. Ask your IT supplier which logs it collects, where they are stored and who can view them. If you do not yet have this information, admitting it is a useful first step, because it shows where to act.
- Record of processing activities updated with the systems holding data
- List of special category or sensitive data
- System access logs with a defined retention period
- Contacts for your legal adviser and your data protection officer
- Contracts with suppliers that process data on your behalf
3. The response steps, in order
When a suspected theft comes to light, the aim is to understand what happened in an orderly way. The playbook must avoid improvised reactions, such as changing everything at once without knowing how the intruder got in. Every step should be recorded, because the reconstruction will be the basis for notification and for later decisions. The choices made at this stage are the ones the consultant will review later, so they deserve careful attention.
Keep two activities separate, as they are often confused: containing the access and assessing the data. The first must start immediately, without waiting for the full picture. The second takes time and needs input from those who know what the systems contain. Mixing them up leads either to blocking everything before understanding what really needs protecting, or to standing still while access is still open. The first sensible step is to decide what to block first, and why.
- Open the action log and appoint an investigation lead.
- Block the suspicious access and isolate the system or account involved, preserving the logs.
- Export and secure the logs before they are overwritten.
- Identify the set of systems and data that may have been read or copied.
- Reconstruct the intruder's route: entry point, accounts used, movements between systems, destinations.
- Establish which categories of data are involved and which people or entities they relate to.
- Involve the legal adviser and the data protection officer in assessing notification.
- Document the decisions taken, based on the information available at the time.
4. Assessing the impact without inventing numbers
The impact assessment answers three questions: which data has left, how many people could be affected, and what harm could result. You do not need an exact figure to start. What you need is an honest description of what you know, what you do not know and what you are verifying. Estimates should not be invented: it is better to say what you do not know.
The risk to people matters more than the number of records. Health, banking and identity data, or data relating to minors, require a different level of attention from a list of business contacts. The consultant can help you classify the risk and decide whether the individuals concerned also need to be told. Document your reasoning so it can be reconstructed: written reasoning is worth more than a hurried answer.
| Type of data | Question to ask | Who can answer |
|---|---|---|
| Identity and contact details | Could they be used for fraud or targeted phishing? | Data owner and consultant |
| Financial or payment data | Could they enable unauthorised transactions? | Finance, bank and consultant |
| Health or special category data | What harm could the person suffer? | Data protection officer and consultant |
| Confidential customer documents | Are there contractual confidentiality obligations? | Management and legal adviser |
5. Notifications: what is triggered and when
If the data involved is personal, the GDPR requires notifying the breach to the supervisory authority within 72 hours of becoming aware of it, unless a risk to individuals is unlikely. If the risk is high, communication to the individuals concerned may also be required, without undue delay. This applies even if discovery is not immediate: the moment you became aware must be recorded.
For entities falling under the European NIS2 Directive, on the security of network and information systems, transposed into Italian law by D.Lgs. 138/2024, the first report of a significant incident must be made within 24 hours, with a more detailed notification within 72 hours. Here too, the assessment is not immediate. The playbook must state who decides, on what information, and who must be consulted before any official communication is sent.
6. How to get started
This week, check whether you can say quickly which systems hold personal data and who has access to them. If the answer is no, the work to do is not yet the playbook: it is the record of processing activities. Choose one system at a time and ask the person responsible to describe its data in a single sentence. Even this shows forgotten systems. A short, up-to-date list is already a good starting point.
Then ask your IT supplier which logs it keeps for your most sensitive systems and for how long they remain available. Write the answer into the playbook. It is the information that will determine, more than anything else, how much you can demonstrate after a theft. If there is no answer, raise the issue with management as a risk to close. A known and recorded risk is easier to govern than an unknown one.
Frequently Asked Questions
How do I know whether data has really been copied rather than just viewed?
It is not always possible to tell with certainty. Access logs, outbound traffic volumes, export operations and unusual file requests are the main indicators. If logs are missing, the assessment must start from the most cautious assumption. A forensic analysis can clarify the point, but it should be entrusted to specialists who have the right skills and preserve the evidence correctly.
Do I need to tell the people whose data has left?
It depends on the risk. For personal data, the GDPR requires communication to the individuals concerned where the breach is likely to result in a high risk to their rights and freedoms, without undue delay. Not every breach requires this step. The decision should be taken with the consultant and the reasoning documented.
How much does the number of records matter in the assessment?
It matters, but not exclusively. Even a limited number of records can have serious consequences if they contain health data, identifiers or credentials. Conversely, a large volume of non-sensitive data may be less significant. For this reason, the playbook should guide the assessment by the nature of the data and the possible harm, not only by quantity.
Is an employee who copied data onto a personal medium an incident?
Yes, it should be treated as one. Not all cases are intentional, and some arise from work habits. The playbook should provide a route for discussion with the staff member and with management, with the legal adviser involved. Technical measures, such as restricting external media, reduce the risk but do not replace the procedure.
Is the supplier responsible for the theft of my data?
Responsibility depends on the contract, the supplier's role and how the theft occurred. A supplier that processes data on your behalf has defined obligations, but you generally remain the controller responsible for the assessment and for decisions towards the authority and the individuals. Check with your consultant what the contract provides and what reporting obligations the supplier has.