An incident without an attacker
A large share of information security incidents do not start with a hacker. They start with an employee who deleted a folder on the file server, a customer spreadsheet sent to the wrong address, or a system administrator who changed a firewall setting and found out the next day that an internal service had been exposed to the internet. Nobody meant to cause harm, and the result can be very similar to that of an attack: lost data, a system that does not work, or personal data that reached someone who was not supposed to see it.
Human error is an unavoidable part of working with systems. You cannot build an organization in which nobody makes mistakes. You can build an organization in which one mistake does not turn into a disaster: permissions limit the damage, there is something to restore from, and whoever made the mistake reports it right away.
The common mistakes
- Deleting or overwriting: a folder deleted on the server or in the cloud, a file saved over a previous version, and a deletion that syncs immediately to every computer.
- Sending to the wrong recipient: address autocomplete that picked a different contact, the wrong file attached to an email, or a customer list sent in visible copy (CC) instead of blind copy (BCC).
- Sharing that is too open: a link to a cloud folder set to “anyone with the link” that keeps circulating among outside parties.
- Misconfiguration: a firewall rule opened for testing and forgotten, cloud storage set to public, two-factor authentication (MFA) turned off “just for a moment” and never turned back on.
- Acting in the wrong environment: a command or script run on the production system instead of the test environment.
- What gets forgotten: a domain or security certificate (SSL certificate) that was not renewed in time, and a website or email that stop working one morning.
A relatively new category is automation. A script, an automation tool or an AI agent given broad permissions can carry out a destructive action at large scale and at speed, such as deleting entire folders or an entire server, without any person approving that specific step.
What raises the risk
- Broad permissions: when every employee can delete any folder, one person’s mistake hurts everyone. Daily work in an admin account also increases the possible damage.
- Changes without a process: without change management there is no rollback plan, and nobody knows what changed and when.
- No test environment: every experiment is done straight on the systems the business runs on.
- Working under pressure: urgent fixes at night, fatigue, multitasking and a deadline tomorrow morning.
- A backup that was never tested: the mistake itself can be fixed, until it turns out the backup did not include the folder or has not run for months.
- Permissive defaults: open sharing by default, no warning about an external recipient and no send delay.
- Knowledge held by one person: only one employee knows how the system is built, and there is no documentation others can work from.
Early warning signs
The signs of the incident itself are usually clear: files that disappeared, employees complaining that a folder is empty, a service that stopped working right after a change, or a customer who replies “I think this reached me by mistake”. Monitoring systems can detect some cases earlier, for example an alert on an unusual volume of deletions or on an external sharing link to a sensitive folder.
Just as important are the signs that the organization is close to a mistake like this: quick fixes done straight on production systems, permissions added “temporarily” and never removed, settings nobody remembers the reason for, and employees who prefer to fix things quietly instead of reporting. Each of these is a warning worth dealing with before the day something breaks.
The impact on the organization
Deleting and overwriting hurt the availability and integrity of data. If there is a good backup, the cost is a few hours of restoring. If there is not, the cost is work that has to be redone, and sometimes data that cannot be recovered at all.
Sending to the wrong recipient and open sharing hurt confidentiality. When it involves personal data, such as customer details, payslips or medical information, this may be a security incident under the Privacy Protection Law and the Privacy Protection Regulations (Data Security), 2017, and in some cases there is a duty to report it to the Privacy Protection Authority. Since Amendment 13 the Authority has broader enforcement powers, and the question that will be asked is what measures the organization took to prevent the mistake and limit the damage.
A misconfiguration is sometimes the most dangerous case of all, because nobody notices it. A service exposed to the internet or cloud storage that became public can stay open for weeks, and there are people who scan the internet precisely to find things like that. That is how an innocent mistake becomes the initial access point for a real attack.
How to reduce the risk
- Least privilege: each employee accesses only what they need, and delete permission on shared folders is granted with care.
- Privileged account management (PAM): administrative actions are done in a dedicated account, not in the account used for reading email and browsing.
- Change management: every change to central systems is documented in advance, with an explanation, a time window and a way to go back (rollback). For critical changes, another person reviews the change before it is carried out (four eyes principle).
- A test environment (staging): significant changes are tested first on a copy, and only then go to the system the business runs on.
- Backup with versions (versioning): the ability to go back to a version of a file or folder from a specific point in time, including a separate backup of Microsoft 365 or Google Workspace, and regular restore tests.
- Protections in email and sharing: a prominent label for external recipients, a short send delay, data leak prevention (DLP) and a sharing policy that limits open links.
- Limits for automation and AI agents: narrow permissions, separation between environments, and a person’s approval for any action that deletes or changes things at large scale.
- Short, focused training: how to send sensitive information safely, and what to do the moment you realize you made a mistake.
Some mistakes happen at the hands of whoever manages the systems. An IT company that works with change management, a rollback plan and a review before execution reduces the risk that it will itself be the source of the incident.
When it happens: stop, report, restore
The first step is to stop the damage and immediately inform whoever manages the systems. Any further action on the affected data may make the restore harder.
- Stop: do not keep working on the folder that was deleted and do not try to fix it yourself. When the deletion syncs to the cloud, whoever manages the systems can stop it from spreading.
- Report: what happened, when, and exactly what was sent or changed. An accurate, fast report saves hours of investigation.
- When sending to the wrong recipient: contact the recipient, ask them to delete the message and confirm in writing that it was deleted. Recalling a message from the email software does not always work, so do not rely on it alone.
- When there is a misconfiguration: return the setting to the last state that worked, and check whether anyone exploited the opening while it was open.
- Restore in a controlled way: restore to a separate location, check that the data is correct and complete, and only then put it back in place.
- Assess the exposure: a cyber incident response team checks what data was exposed and to whom, and whether a report to the Privacy Protection Authority or a notice to the people whose data was sent is required.
- Analyze the root cause (root cause analysis): the central question is what in the system or process allowed this mistake to cause this kind of damage, and less what the employee did.
A culture that makes reporting possible
One of the factors that most affects the size of the damage is the time that passes until someone reports. An employee who fears the reaction will try to fix it quietly, wait in the hope that nobody notices, or erase the traces. Meanwhile the old backups are replaced by new ones, the wrong email is forwarded on and the opening in the firewall stays open.
Organizations that handle mistakes without looking for someone to blame (blameless culture) get earlier reports and learn from them. That does not mean there is no accountability. It means the first question is how to fix it, and the second is what to change so it does not happen again. A simple reporting procedure everyone knows, with one address or one number to contact, makes a big difference here.
Summary
- Permissions on shared folders and systems are limited by role.
- Administrative actions are done in separate accounts, and central changes are documented with a way to go back.
- There is a backup with versions, including Microsoft 365 or Google Workspace, and a restore was tested recently.
- Email labels external recipients, and the sharing policy limits open links.
- Automations and AI agents have limited permissions and require a person’s approval for destructive actions.
- Employees know who to report to, and know that reporting quickly is the right thing to do.