Canadian data residency · made, hosted, and operated in Canada

Early access — 90 days →

MITRE ATT&CK Enterprise · TA0040

Impact: the damage and the extortion

ImpactTA0040

Everything before this was reconnaissance. This is the part that costs money. What destruction, encryption and denial of service look like from the outside, and how the telemetry catches them.

Codein Cybersecurity November 22, 2026 7 min read

For the owner

Every intrusion has a moment when it stops being a security problem and becomes a business problem. That moment is not the break-in. It is the point at which files stop opening, servers stop answering, or a page on your site explains the price of reaching you.

The range of damage is wider than ransomware. An intruder with access to your systems can delete the backups and then encrypt everything — the same event with a ransom note attached. They can stop a service your business runs on. They can delete a domain account, including yours, and lock the company out of its own systems. They can change a stored value — a bank account, a delivery address — and let the business act on it for weeks.

Each looks different from the outside. Ransomware is loud: people notice in minutes. A manipulated supplier bank account is silent, and is discovered when the money does not arrive.

Three questions worth asking your own team:

  1. If your backups were deleted tonight, would the business survive? Restoring is not the same as having backups. Have you watched one happen?
  2. Who can stop the systems the business depends on, and who can delete the accounts that stop them?
  3. What does one hour of downtime cost you, and one lost week of payroll data?

The honest answer is that nobody has checked since the last change of system. That is the gap this tactic exposes.

For the analyst

The tactic

MITRE ATT&CK files Impact under TA0040. It is the last tactic in the chain and the one with the highest ratio of consequence to bytes: a single event here is worth more than a thousand events in an earlier tactic. The matrix files 33 techniques under it, and the pinned corpus carries 103 rules that name a technique in this tactic.

Impact is where the corpus is densest, and the reason is visible in the table: the most-covered techniques are the ones an operator performs deliberately and with command lines. T1490, inhibiting system recovery, is the largest block in the tactic, because deleting a backup leaves a command line that names a backup tool.

TechniqueNameRules in the corpus
T1490Inhibit System Recovery23
T1485Data Destruction19
T1489Service Stop19
T1496Resource Hijacking13
T1486Data Encrypted for Impact10
T1531Account Access Removal9
T1529System Shutdown/Reboot8
T1565.001Data Manipulation: Stored Data Manipulation6
T1491.001Defacement: Internal Defacement4
T1499.004Endpoint Denial of Service: Application or System Exploitation3
T1565Data Manipulation3
T1498Network Denial of Service2
T1565.002Data Manipulation: Transmitted Data Manipulation2
T1495Firmware Corruption1
T1499Endpoint Denial of Service1
T1499.001Endpoint Denial of Service: OS Exhaustion Flood1
T1561.001Disk Wipe: Disk Content Wipe1
T1561.002Disk Wipe: Disk Structure Wipe1

The order of the attack

Ransomware does not begin with encryption. It begins by removing the ability to recover. The ATT&CK ordering here is also the operational ordering:

  1. Remove the recovery path (T1490). Shadow copies deleted, backup catalogues emptied, system restore disabled, boot configuration tampered with.
  2. Remove the services and the accounts (T1489, T1531). Backup services stopped, database services stopped, the intruder account and sometimes the administrator's account removed or locked.
  3. Encrypt or destroy (T1486, T1485).
  4. Occasionally, stage a scene (T1529, T1491.001, T1565).

Every one of those steps has an innocent explanation that an administrator will give you at 2 a.m. Only the sequence is an attack, which is why this tactic cannot be detected by watching one event.

The techniques, and what carries them

Inhibit System Recovery (T1490). The corpus is specific: wbadmin deleting a backup catalogue, bcdedit boot configuration tampering, reagentc disabling the recovery environment, registry writes that disable system restore, PowerShell that calls the shadow copy API, tmutil on macOS, and cloud API calls that disable versioning on a storage bucket. These are near-zero false positive events in any environment that has no business running them at that hour.

Data Encrypted for Impact (T1486). A file rename event carrying a new extension, a Microsoft 365 rule for potential ransomware activity, a BitLocker registry change, a cloud KMS import of attacker key material. The encryption itself is usually the last event, not the first.

Service Stop (T1489). net stop, sc stop, PowerShell Stop-Service, schtasks /delete, and on Linux systemctl stop --now seen through auditd. The signal is rarely the stop itself but the parent: a service stopped by a process that is not the service manager, or a stop of a service that has never been stopped before.

Data Destruction (T1485). dd writing over files, cipher /w, Sysinternals sdelete, fsutil against a volume, a destructive SQL statement, and a cloud rule for an unusual volume of file deletion.

Account Access Removal (T1531). userdel, groupdel, removal of a user from an administrative group, a Google Cloud service account disabled, an Okta user locked out. On its own this looks like offboarding, which is precisely why it is worth an alert: the intruder is removing the people who would notice.

System Shutdown/Reboot (T1529). shutdown, a Linux poweroff, a virtual machine killed from the hypervisor, an auditd shutdown record.

Data Manipulation (T1565, T1565.001, T1565.002). The quiet one. Stored data manipulation covers cleared system logs, deleted shell history and altered configuration; transmitted data manipulation covers the value in flight. Neither makes a sound. This is the technique that turns a supplier bank account into an incident six weeks later.

Defacement (T1491.001). A changed desktop background, a registry desktop-background write, a legal notice message. Four rules in the corpus: easy to detect, rarely material financially — unless it is aimed at your customers.

Resource Hijacking (T1496). Cryptocurrency mining: mining pool domains in DNS, miner process names, cloud audit events where clusters are created to mine. The damage is the electricity bill and the capacity.

Denial of service (T1499, T1499.001, T1499.004, T1498). Endpoint exhaustion floods, application or system exploitation used to crash a host, and network floods. A separate family from destruction: the system survives and stays down.

Firmware corruption (T1495) and disk wipe (T1561.001, T1561.002). The irreversibility cases — a corrupted BIOS, a wiped volume structure. One rule each, because they are rare, and rare is what makes a rule worth writing.

How a rule reasons over this telemetry

The normalized columns a rule sees here are process_name, parent_process, command_line, file_path, user_name, host_id, event_type and severity, and the sources are windows_sysmon, linux_auditd, the Windows security and service-control logs, and cloud audit trails.

A good rule in this tactic reasons about three things rather than one:

  • Who. The user_name and host_id. A wbadmin run by a service account on a schedule is normal; the same binary run by a user session on a workstation is not.
  • When. The time of day relative to the machine's own rhythm, and the sequence relative to other events on the same host.
  • What came before. The highest-value context for T1490 is the preceding minutes of that host's history: was there a credential dump, an unusual sign-in, a new service installed? The rule that fires "shadow copy deletion, and the last 15 minutes contain a credential dump" is far better than the rule that fires on shadow copy deletion alone.

What makes a rule wrong is the administrative baseline. Backups, disk maintenance, password rotation for a departing employee, a database rebuild — each of these produces the same command lines as an attack. The discriminating questions are always which account, which host, which hour, and was this account ever seen doing this before.

Detection engineering notes

  • Alert on the first irreversible step, not the loudest one. The encryption is loud and late. The backup deletion is quiet and last-chance. Weight the recovery-inhibition rules accordingly: they are rare, so they can afford to be less aggressively suppressed than a noisy technique.
  • Correlate on host_id inside a time window, not alert by alert. The sequence is the signal. A correlation window of tens of minutes across the host's events is what turns T1490 plus T1489 plus T1486 into one incident instead of three tickets.
  • Bind to parent_process and user_name, and treat command_line as supporting evidence. Attackers change binary paths; they do not change who is logged in.
  • Keep the quiet rules. Manipulation and defacement will never make your alert volume interesting, and they are exactly the findings that a customer would later say you missed.
  • Decide in advance what happens when an impact alert fires. Detection without a rehearsed response is a record of the incident, not a way to avoid it.

Where analysts get it wrong

The classic false negative is encryption that happens on a file server. If the intruder encrypts a share from the server, there are no endpoint file events at all — the signal is in the file server's own logs and in the network session, and a rule that needs file_path from a workstation will never fire.

The second false negative is the patient intruder. A dwell time of a month means the reconnaissance never triggers anything. Nothing in a rule set based on "first seen" behaviour helps, because by then the host's baseline includes the intruder.

The third is the extortion email that arrives before the encryption. Some operators steal first and threaten later, which means the only alert you get is a phishing detection on the morning the notice lands, and the exfiltration alert came weeks earlier and was filed as a false positive.

The fourth is treating "impact" as if it only means ransomware. T1565 and T1565.002 are the findings that cost a business real money without producing a single alert, and T1496 — a miner quietly running on a workstation for two months — looks exactly like a slow server until somebody looks at the electricity bill.

Where DetectMonkey fits

At this point in the chain, detection latency stops being a technical metric and becomes a business one. The gap between the backup deletion and the encrypted file is the gap between a Tuesday morning and a Tuesday afternoon.

DetectMonkey watches that gap. The collectors read the process creation, file, service and registry events this tactic produces, and the pinned corpus of 3164 rules evaluates them as the events arrive, so the quiet step — the one that has no user-visible symptom — is the step that gets alerted on.

The platform detects and alerts. It does not isolate, quarantine or block anything on your systems, it does not remediate anything, and it changes no configuration of yours. What it does is put the sequence in front of a human, with the evidence attached, in time to do the thing that actually saves the business — decided before the encryption starts.

The telemetry is processed and stored on private servers operated by Codein in New Brunswick, Canada. Analysts work in English and French.

Early access is 90 days at no charge, for up to 100 seats, with founder-led onboarding and no card required.

See it on your own telemetry

Early access is 90 days at no charge, for up to 100 seats, founder-led onboarding, no card required.