Partition-table wipe utility invoked on Linux

Siemphony@siemphonylowT1561.002unverified

Implements the tool-invocation leg of AN0828 on auditd: execution of one of three non-interactive utilities capable of destroying a partition table in a single command — sgdisk, sfdisk, wipefs. The device-path leg that used to sit alongside it has been removed. T1561.001 is already authored in this corpus on the identical logsource with a device selector (`name|startswith` on `/dev/sd`, `/dev/nvme`, `/dev/vd`, `/dev/xvd`, `/dev/hd`, `/dev/mmcblk`, `/dev/md`, `/dev/dm-`, `/dev/mapper/` and `/dev/disk/`) that is a strict superset of the fifteen whole-disk nodes this rule used to enumerate, so every device-leg match here also raised that sibling rule and nothing on the event told the two techniques apart — a write to /dev/sda is equally consistent with a content wipe (.001) and a structure wipe (.002), and the offset that would distinguish them is not a field auditd exposes to Sigma. Keeping the device leg added alert volume, not discriminating power, so it is dropped and this rule now stands only on the coverage genuinely specific to Disk Structure Wipe: sgdisk, sfdisk and wipefs, whose purpose is the partition table rather than disk content, and which are disjoint from T1561.001's dd/shred/wipe/nwipe/scrub/blkdiscard/badblocks tool list. `dd`, `fdisk`, `gdisk`, `parted` and `cfdisk` stay excluded, as before: `dd` is too generic to select on its own name, and the interactive tools cannot be separated from provisioning work without arguments this feed does not supply on the same record as the binary. This leg has no argument gate, which is a real limitation and not just an unused knob: MITRE's log source for AN0828 names arguments as part of the observable ("writing to sector 0 or partition table"), but a raw auditd stream splits one execution into a SYSCALL record carrying `exe` and a separate EXECVE record carrying the arguments, so this selection — matching on `exe` alone — fires identically on a destructive call and on a read-only one. `sfdisk -l` and `sfdisk -d` dump the layout, `sgdisk -p` and `sgdisk --print` print the table, and bare `wipefs /dev/sda` only lists signatures — only `wipefs -a` erases them — and none of that is distinguished here. MITRE's `OffsetThreshold` knob is unreachable for the same reason a write predicate is: this leg observes process launch, never the bytes a syscall touched. The Windows arm AN0827 is not attempted, because Security 4673 carries only `SubjectUserName`, `ProcessName`, `PrivilegeList` and `Service`, and Sysmon `driver_load` carries only `ImageLoaded`, `Hashes` and `Signature` — neither has a field that could hold a `\.\PhysicalDrive` path. Prerequisite: auditd records no execve at all until an exec rule such as `-a always,exit -F arch=b64 -S execve -k exec` is loaded, with its b32 twin — without it this rule returns zero rows, and zero rows here reads as quiet when it actually means blind. UNVERIFIED AS AUTHORED — derived from MITRE ATT&CK DET0297, not hand-written and not tested by its author. Any lab result is on this rule's own page.Full description

The 3 SIEM queries below are previews produced by Siemphony’s own translator, not pySigma, and none has been executed against a real backend — review before you deploy.

detection.yml

Splunk · SPL

Run this as a search.

index=* (type="SYSCALL" AND (exe="*/sgdisk" OR exe="*/sfdisk" OR exe="*/wipefs"))

Elastic · ES|QL

Run this as a search.

FROM logs-*| WHERE (TO_LOWER(type) == "syscall" AND (TO_LOWER(exe) LIKE "*/sgdisk" OR TO_LOWER(exe) LIKE "*/sfdisk" OR TO_LOWER(exe) LIKE "*/wipefs"))

Wazuh · XML rule

Deploy to your manager — this is a rule, not a search.

<group name="sigma,linux,">  <!-- Rule ids must be unique on your manager; 100000+ is the user range. -->  <rule id="100000" level="5">    <!-- Set <if_sid> to the decoder/base rule for linux so this only evaluates relevant events. -->    <field name="type" type="pcre2">(?i)^SYSCALL$</field>    <field name="exe" type="pcre2">(?i)(/sgdisk$|/sfdisk$|/wipefs$)</field>    <description>Partition-table wipe utility invoked on Linux</description>    <mitre>      <id>T1561.002</id>    </mitre>  </rule></group>

Sentinel · KQL

Run this as a search.

Sentinel cannot express this construct. No Sentinel table is modelled for this logsource ({"product":"linux","service":"auditd"}), so there is nothing honest to query. Pick a logsource this translator maps, or write the KQL by hand. The Sigma source is on the first tab, unchanged.

At a glance

Logsource: auditd / linux
Level: low
Category: impact_destruction
Sigma status: experimental
Backends: 3 of 4 rendered
Updated: 30 Aug 2026
Rule ID: 5e251b45-f428-4164-bff9-f247bdb0feb7

ATT&CKT1561.002

T1561.002Disk Structure Wipe

ATT&CK® is a registered trademark of The MITRE Corporation. These mappings are the rule author’s, and are neither reviewed nor endorsed by MITRE.

Lessonimpact destruction

impact destruction

01 · What this rule detects

The work an operator does to make recovery impossible before, or while, destroying data — deleting volume shadow copies, wiping backup catalogs, disabling Windows recovery and boot repair, and deleting or overwriting files recursively. Most of it happens in a short burst immediately before encryption, which is why the precursors are the part worth detecting: by the time file modification volume is the signal, the decision has already been made for you. This overlaps deliberately with log clearing — the same burst usually destroys evidence and recovery together — but keep the two rules apart, because they point at different questions.

02 · Telemetry you need

Process creation with full command lines — Sysmon EventID 1, or Security 4688 with Audit Process Creation and the command-line inclusion policy; without the second policy you see that vssadmin ran and nothing about what it did
The utilities themselves: vssadmin, wmic shadowcopy, wbadmin, bcdedit, cipher, and the PowerShell and WMI equivalents that never touch a command-line binary
Sysmon EventID 23 and 26 (FileDelete and FileDeleteDetected) — not in the default configuration, and the archiving variant needs disk space budgeted for it
Service and provider activity for the Volume Shadow Copy service, and the backup product's own logs, which record deletion the operating system does not
Security 4663 for delete access on protected shares — Audit Object Access plus a SACL, off by default and inert without the SACL

03 · How the logic works

These command lines are short, fixed and close to unique, which makes this one of the few categories where string matching against the command line is the right primary approach rather than a fallback. Structure each as image AND destructive verb: matching vssadmin alone catches every backup product listing shadows, whereas vssadmin with delete does not. Allow for argument reordering, abbreviation and the several spellings of the same operation, and remember that the same effect is reachable through WMI and the COM interfaces without running the utility at all — so a rule set that names only the executables has a hole in it that any competent tooling walks through. The mass-encryption half is a rate problem: file modifications per process per minute, which Sigma cannot express and which belongs in the correlation layer.

04 · Common false positives

Backup software managing its own shadow storage and pruning old copies — Veeam, Acronis, Windows Server Backup and the built-in scheduled tasks
Disk cleanup and storage reclamation removing shadow copies when a volume fills
Imaging, provisioning and task sequences that use bcdedit and reset recovery configuration by design
Non-persistent VDI and lab rebuild workflows that discard machine state on every cycle
Developers and build systems recursively deleting output directories

05 · Detection maturity

Tier 1 — starting point: Matches vssadmin and wbadmin deletion command lines.
Tier 2 — tuned: Covers the WMI and PowerShell routes and boot configuration tampering, with backup hosts and maintenance windows excluded.
Tier 3 — correlated: Correlates recovery destruction with log clearing, service stops and file modification rate from the same host and process tree.

Tuning & false positivesexperimental · 5 declared

experimental (current stage)
test
stable

Written and reasoned about, but not yet run against real telemetry. Expect to tune it before trusting it.

Tuning tips

This should be a near-silent category. If it is noisy, find out what is deleting recovery points routinely and fix that rather than filtering it
Never exclude by account — the accounts used here are administrative by definition, which is the point
Scope exclusions to the specific backup hosts and their maintenance windows, and revisit them when the backup product changes
Alert on failed attempts as well as successful ones; a denied shadow deletion is the same intent and better news
Escalate hard on the combination — shadow deletion, log clearing and service stops from one host inside a few minutes is a different alert from any of them alone

Declared by the author

Cloud and virtual-machine first boot. cloud-init's growpart module shells out to sfdisk for an MBR-labelled disk or sgdisk for a GPT-labelled one to expand the root partition on the first boot of every instance built from an image, so on any estate that provisions VMs this is the highest-volume match by a wide margin and it is what sets the level.
Software-defined storage preparing disks. ceph-volume and Rook run wipefs -a and sgdisk --zap-all against every candidate device before creating an OSD, and Kubernetes local-storage and LVM provisioners do the same, so a storage node emits this rule on every disk add, OSD rebuild or node reimage.
Operating-system installation and imaging. Anaconda with a kickstart, debian-installer preseed, Ignition and Foreman all call sgdisk or sfdisk to lay out the partition table during an unattended build, producing a burst of matches for every host provisioned.
Configuration management running disk tasks on a schedule. Ansible community.general.parted, the filesystem module, and hand-written wipefs or sfdisk tasks re-assert the layout of attached data disks on every run, whether or not anything has changed.
Inventory, fact-gathering and monitoring scripts that only read the partition table: sfdisk -l or sfdisk -d, sgdisk -p or sgdisk --print, and a bare wipefs invocation (no -a) all list rather than modify, but the argument that would tell a listing apart from a wipe lives on a separate auditd record this rule does not read, so CMDB collectors and Ansible fact-gathering polling on a schedule match identically to a destructive call.

Noise check0 of 5 fired

Fires on none of the 5 ordinary behaviours the baseline holds for this log type. That is not a precision guarantee: it means only not caught by these 5 behaviours, and the next ordinary behaviour nobody wrote down may still match.

Run live against 65 hand-written events of ordinary corporate activity — backup agents, software distribution, patching, helpdesk scripts, EDR. It says nothing about how often the rule would fire in production, which is the number that actually exhausts an analyst.

Verification1 run

Not verified

Siemphony benign baseline
completed
Siemphony IR · 3 Sept 2026
No events matched, so this run proves nothing either way.

Test it yourselfin your browser

Runs here in your browser against public attack recordings. A preview — not the verification badge.

Recordings of T1561.002, evaluated against their — events.

Share this rule

/u/siemphony/5e251b45-f428-4164-bff9-f247bdb0feb7

More from this repertoire

All compositions

Verdicts · reactions · comments

Community

Verdicts from engineers who actually deployed it, and the conversation around it.

Not yet reported on

Nobody has run this in a real environment and said what happened.

Verdicts from engineers who deployed it

0 cast

No verdicts reported yet. The first one is worth more than the tenth — say what you ran it against.

No comments yet. Someone who deploys this will have something to say about it.

Partition-table wipe utility invoked on Linux

The detection

detection.yml

Splunk · SPL

Elastic · ES|QL

Wazuh · XML rule

Sentinel · KQL

Community

Verdicts from engineers who deployed it