>
av.Amit VijayanCYBERSECURITY & RESEARCHLet\u2019s connect
Back to the library

Threat Hunting Playbooks

Hunting CVE-2024-10441 Exploitation: Synology DSM Unauthenticated RCE in Splunk and CrowdStrike

Thumbnail for: Hunting CVE-2024-10441 Exploitation: Synology DSM Unauthenticated RCE in Splunk and CrowdStrike

Why this matters now

On November 18, 2026, CISA added CVE-2024-10441 to its Known Exploited Vulnerabilities catalog, with a federal remediation deadline of November 21 — tomorrow. The flaw is an unauthenticated remote code execution bug in the system plugin daemon of Synology DiskStation Manager (DSM) and BeeStation OS (BSM): a remote attacker needs no login at all to run code on the NAS. The patch has existed since December 2024 — Synology disclosed the flaw in November 2024 and published the details in March 2025 — and yet the vulnerability is being exploited now, two years later, because unpatched devices are still reachable. On October 29, 2026, Synology said threat actors were exploiting a vulnerability patched in December 2024 to target devices running outdated DSM, and that the attacks aim to encrypt data and extort ransom — the C3RB3R ransomware campaign against NAS devices. The deadline is three days from the KEV listing because this is the most severe tier under CISA's BOD 26-04: unauthenticated, network-reachable, fully automatable, and handing the attacker full control of the system. CISA also ordered forensic triage for this entry, which means patching is not the whole job — you have to determine whether the device was already entered.

CVECVE-2024-10441 — Improper Encoding or Escaping of Output in the system plugin daemon of Synology DiskStation Manager (DSM) and BeeStation OS (BSM)
CVSS9.8 Critical (CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H), scored by Synology as the CNA and published identically on the NVD entry. One disagreement to note: the Zero Day Initiative advisory ZDI-25-214, which covers the same flaw on a DiskStation DS1823xs+, scores it 8.8 with an adjacent-network vector (AV:A) rather than network (AV:N). This article leads with the vendor's 9.8 per the standing rule — Synology rates its own product as remotely reachable — and flags the ZDI difference because it matters for exposure: if your NAS is not directly internet-reachable, the ZDI framing is the one that applies to you.
WeaknessCWE-116 — Improper Encoding or Escaping of Output, per NVD and Synology
Affected productSynology DiskStation Manager (DSM) before 7.2-64570-4, 7.2.1-69057-6, 7.2.2-72806-1, 7.1.1-42962-7 and 6.2.4-25556-8; BeeStation OS (BSM) before 1.1-65374; DSMUC 3.1 before 3.1.4-23079
VendorSynology
DisclosureAdvisory Synology-SA-24:20 (DSM) and Synology-SA-24:23 (BSM) first published November 5, 2024; vulnerability details disclosed March 19, 2025. The flaw was reported through Pwn2Own 2024 as ZDI-CAN-25403, credited to Pumpkin Chang and Orange Tsai of DEVCORE and to Ryan Emmons of Rapid7.
CISA KEVAdded November 18, 2026; federal remediation due November 21, 2026. Known ransomware campaign use: Unknown. Forensic triage: required under BOD 26-04.
Fixed inDSM 7.2.2-72806-1, 7.2.1-69057-6, 7.2-64570-4, 7.1.1-42962-7, 6.2.4-25556-8; BSM 1.1-65374; DSMUC 3.1.4-23079. Synology lists no mitigation — the only fix is the update.
ExploitationConfirmed — CISA added the flaw to KEV on evidence of active exploitation, and Synology confirmed on October 29, 2026 that threat actors are exploiting a December-2024-patched DSM flaw against outdated devices in ransomware attacks. One honest limit: neither CISA nor Synology has said CVE-2024-10441 is the specific flaw used in the C3RB3R campaign — Synology pointed at advisory SA-24:20 as a whole, which covers three CVEs, without naming one. Separate reporting links C3RB3R to two other DSM flaws (CVE-2026-13684 and CVE-2026-13639), which are not in KEV. What is established is that this CVE is being exploited in the wild; which ransomware operation used it, and at what scale, is not publicly established.

The hunting window is long. The patch shipped in December 2024, the details went public in March 2025, and exploitation is confirmed in November 2026 — so run these hunts back at least 30 days, and treat any device still on a vulnerable build as a device that has been in the target population for two years. The same edge-appliance lesson from the Barracuda ESG hunt applies here — an unauthenticated flaw on the box, a patch that does not evict an attacker who already got in — and the Check Point Gaia hunt shows the same pattern from the other side: the fix closes the hole, the hunt finds who walked through it. This flaw was also covered in the November 19 brief, which has the full KEV context.

The technique, in hunt terms

The flaw sits in the system plugin daemon — the DSM component that handles plugin operations on the NAS. Synology's advisory describes it only as an improper encoding or escaping of output that lets a remote attacker execute arbitrary code via unspecified vectors, and neither Synology nor NVD publishes exploit request syntax. The Zero Day Initiative advisory adds the one mechanical detail that makes this huntable: the flaw is in login handling. The username supplied at login is not validated before it is used to build environment variables, so a crafted username can break out of the intended value and execute code as root. That is why Hunt 1 looks at login events, not at generic web traffic: the exploit, as far as the public record describes it, arrives as a login attempt carrying shell syntax in the username field.

Two companion flaws ship in the same advisory and matter for triage. CVE-2024-50629 (CVSS 5.3) is the same class of weakness in the webapi component and lets an unauthenticated attacker read limited files — a file-read primitive that pairs naturally with the RCE. CVE-2024-10445 (CVSS 4.3, adjacent network) is an improper certificate validation flaw in the update functionality that lets an adjacent attacker write limited files. Neither is in KEV, but a device exploited through CVE-2024-10441 may also show the other two being probed, and a file read through CVE-2024-50629 is itself evidence of unauthenticated access worth hunting. If you find exploitation of the critical flaw, check for the companions too.

In MITRE ATT&CK terms, the exploitation maps to T1190 (Exploit Public-Facing Application) for the unauthenticated entry through the DSM web interface, and T1059.004 (Unix Shell) for the code execution that follows — the NAS is a Linux system, and the attacker lands in a shell as root. The companion file read maps to T1005 (Data from Local System), and the ransomware outcome Synology describes maps to T1486 (Data Encrypted for Impact), labelled here as the campaign outcome rather than a confirmed effect of this specific CVE. Post-exploitation persistence on a NAS typically means a scheduled task (T1053, Scheduled Task/Job — DSM Task Scheduler runs on cron underneath) or a new administrative account (T1136, Create Account), so both are hunted below as hypotheses, not as confirmed campaign behavior.

One honest framing note before the queries: the CrowdStrike Falcon sensor does not run on DSM, so there is no Falcon process data from the NAS itself. The CrowdStrike sets below cover the two places Falcon does see this campaign — the same DSM and firewall logs forwarded into Falcon LogScale as third-party data, and Falcon endpoint telemetry on the workstations and servers that mount the NAS shares, which is where the ransomware follow-on becomes visible. If you do not forward DSM's Log Center logs anywhere today, Hunt 1 is a reason to start: DSM can send its logs to a syslog server, and the login events are the only record of the exploit arriving.

Hunt 1: The login vector — usernames carrying shell metacharacters

The exploit's public description points at one place: the username field of a login attempt. A legitimate username never contains $(, ${, a backtick, a semicolon, a pipe, or &&. DSM's Log Center records login attempts — including failed ones, with the attempted username and the source address — and forwards them to a syslog server when configured. This search looks for exactly those characters in the attempted username. A single hit is not a lead; it is an exploitation attempt against the system plugin daemon, and it deserves the same treatment as a confirmed shell. Run it back 30 days at minimum. Replace the sourcetype with whatever your DSM syslog ingestion names it.

index=* sourcetype="synology:dsm:syslog" earliest=-30d
| rex field=_raw "User \\[(?<attempted_user>[^\\]]+)\\]"
| rex field=_raw "from \\[(?<src_ip>[0-9a-fA-F\\.:]+)\\]"
| where match(attempted_user, "(\\$|`|;|\\||&&)")
| table _time, host, src_ip, attempted_user, _raw
| sort _time desc

The LogScale equivalent, for the same forwarded log ingested as third-party data — adjust the raw-event field name to how the log lands in your repo. LogScale's bare-regex filter does the same metacharacter test in one line:

sourcetype="synology:dsm:syslog" "failed to log in" /User \[.*(\$|`|;|\|)/
| table([@timestamp, host, message])
| sort(@timestamp, order=desc)

Two things to check on every hit. First, whether the same source address also appears in successful logins or in any other DSM event around the same time — an attempt followed by a successful login from the same address is the escalation, not the probe. Second, whether the username contains a recognizable command rather than just a delimiter: id, whoami, curl, wget, nc, /dev/tcp, base64. Those strings inside a username field are the exploit payload, and they are worth a second, narrower pass over the same log:

index=* sourcetype="synology:dsm:syslog" earliest=-30d
  (" id" OR "whoami" OR "/etc/shadow" OR "curl " OR "wget " OR "nc " OR "/dev/tcp" OR "base64" OR "chmod" OR "useradd")
| table _time, host, _raw
| sort _time desc

Hunt 2: The exploitation burst — failed-login storms and web-interface probing

Not every exploitation attempt carries a visible payload in the username, and the KEV listing means automated exploitation is running at scale. The second signal is volume: a burst of failed logins against the DSM web interface from a single source, or a single source trying many distinct usernames. Normal DSM use does not produce twenty failed logins in ten minutes from one address. This search buckets failed logins and surfaces the storms; a source that also appears in Hunt 1 is a confirmed attacker, and a source that only appears here is a scanner worth blocking and watching.

index=* sourcetype="synology:dsm:syslog" earliest=-14d "failed to log in"
| rex field=_raw "from \\[(?<src_ip>[0-9a-fA-F\\.:]+)\\]"
| bucket _time span=10m
| stats count as failures by _time, host, src_ip
| where failures > 20
| sort - failures

The companion view — distinct usernames per source — separates exploit probing from password spraying. A spray tries many passwords against a few real accounts; exploit probing tries many crafted usernames, most of which are not real accounts at all:

index=* sourcetype="synology:dsm:syslog" earliest=-14d "failed to log in"
| rex field=_raw "User \\[(?<attempted_user>[^\\]]+)\\].*from \\[(?<src_ip>[0-9a-fA-F\\.:]+)\\]"
| stats dc(attempted_user) as distinct_users, count as attempts by host, src_ip
| where distinct_users > 10
| sort - distinct_users

On the LogScale side, the same burst view over the forwarded DSM log, grouped by host so a storm against any single NAS stands out:

sourcetype="synology:dsm:syslog" "failed to log in"
| groupBy([host], function=count(as=failures))
| sort(failures, order=desc)

Hunt 3: Post-exploitation on the NAS — shells, outbound connections, and new persistence

Code execution through this flaw runs as root on the NAS, so the post-exploitation evidence is whatever a root shell does next: spawn a downloader, open an outbound connection, create an account, or schedule a task. The NAS itself is the blind spot — there is no endpoint agent on it — so this hunt works from the logs around it. First, the DSM web server's own access log, on the device at /var/log/nginx/access.log and forwardable the same way as the Log Center logs: requests to the DSM web interface carrying URL-encoded shell syntax are the exploit arriving over HTTP, and they are visible even when the Log Center event is not:

index=* sourcetype="synology:nginx:access" earliest=-30d
| search ("$%28" OR "%24%28" OR "%24%7B" OR "%60" OR "%3B" OR "%7C" OR "$(" OR "${")
| table _time, host, clientip, uri_path, uri_query, status
| sort _time desc

Second, the network view. A NAS is a storage box: its normal outbound traffic is updates, NTP, and the services you configured. A newly seen outbound destination from a NAS IP — especially on an unusual port, or to an address with no history — is the shell calling home or pulling a payload. Replace the placeholder IPs with your NAS addresses, and run the window back 30 days:

index=network earliest=-30d src_ip IN ("<nas-ip-1>", "<nas-ip-2>")
| stats earliest(_time) as first_seen, count as connections by src_ip, dest_ip, dest_port
| where first_seen > relative_time(now(), "-14d")
| sort first_seen desc

The LogScale equivalent over ingested firewall or proxy logs, grouped by destination so a new call-home address stands out:

#repo=firewall src_ip="<nas-ip>"
| groupBy([dest_ip, dest_port], function=count(as=connections))
| sort(connections, order=desc)

Third, persistence on the device. DSM logs account and scheduled-task changes in Log Center, so search the forwarded log for new users, privilege changes, and Task Scheduler activity. DSM log wording varies by version, so treat these strings as a starting point and adjust them to your build — a new administrator account or a new scheduled task created in the last 30 days on a NAS that nobody administers by hand is the finding, whatever the exact wording:

index=* sourcetype="synology:dsm:syslog" earliest=-30d
| search ("Task Scheduler" OR "created" OR "administrators")
| table _time, host, _raw
| sort _time desc

Hunt 4: The ransomware follow-on — mass encryption on mapped shares

The outcome Synology describes is encryption of the data on the NAS, and that is where Falcon endpoint telemetry finally sees the campaign. The ransomware runs on the NAS, but it encrypts the shared folders — and every workstation and server that mounts those shares sees the file churn. A single process writing an unusually large number of files in a short window, or a process spawning shell commands and downloaders on a host that mounts NAS shares, is the follow-on. This LogScale query surfaces processes whose command lines carry the classic post-exploitation tooling — reverse shells, downloaders, and interactive shells — on any host in the estate:

#event_simpleName=ProcessRollup2
| CommandLine=/curl|wget|nc |ncat|\/dev\/tcp|mkfifo|bash -i|sh -i/
| table([@timestamp, ComputerName, UserName, ImageFileName, CommandLine])
| sort(@timestamp, order=desc)

And the volume view: count file-write events per host and look for the outlier. Normal hosts do not suddenly write ten thousand files in an hour; a host encrypting a mapped share does. Baseline your estate first — backup software and build systems are the usual false positives — then investigate the host that breaks the baseline, and check whether it mounts a share on an unpatched NAS:

#event_simpleName=FileWritten earliest=-7d
| groupBy([ComputerName], function=count(as=file_writes))
| file_writes > 10000
| sort(file_writes, order=desc)

On the Splunk side, the same follow-on is visible in the file server or endpoint data you already ingest: a burst of file modifications across a mapped share, or ransom-note files appearing in shared folders. Treat any ransom note as confirmed compromise of that share's data, and scope it by asking which NAS hosts the share and whether that NAS is still on a vulnerable DSM build.

Triage: sorting real compromise from noise

A hit in Hunt 1 — shell syntax in a username — is an exploitation attempt, not noise, and it goes straight to incident response. A hit in Hunt 2 alone is usually scanning: block the source at the perimeter, check whether the same source appears in Hunt 1 or Hunt 3, and move on if it does not. The question that decides the severity of any hit is the DSM build: check Control Panel, Update & Restore on the device, or cat /etc/VERSION on the NAS itself, and compare against the fixed builds — 7.2.2-72806-1, 7.2.1-69057-6, 7.2-64570-4, 7.1.1-42962-7, 6.2.4-25556-8, or BSM 1.1-65374. A vulnerable build with a Hunt 1 hit is a presumed compromise until proven otherwise; a patched build with the same hit is a blocked attempt, and the log becomes evidence the patch is doing its job.

For a presumed compromise, the triage order is: first, confirm the build and the patch state, and check the device for new administrator accounts, new scheduled tasks, and unfamiliar processes — the persistence from Hunt 3. Second, check the companion flaws: look for unauthenticated file reads through the webapi component (CVE-2024-50629) in the same window, since a file read is evidence the attacker was already inside. Third, check the shares: look for mass file modification, renamed files, or ransom notes, and check your backups before you touch anything — restoring from a backup that the attacker also encrypted is the second incident. Fourth, remember that CISA requires forensic triage for this entry under BOD 26-04: preserve the DSM logs, the nginx access log, and the firewall logs for the window before you reimage or restore, because the exploitation evidence is the record of what the attacker reached.

One caution on attribution: do not conclude from a Hunt 1 hit that the device was hit by the C3RB3R campaign specifically. CISA's KEV entry lists ransomware use as Unknown, and the public record does not tie C3RB3R to this CVE. The hit proves exploitation of CVE-2024-10441; the campaign behind it is a separate question, and the answer does not change the response.

Remediation and hardening

Patch first, and patch to the build for your branch — DSM 7.2.2-72806-1 or later, 7.2.1-69057-6 or later, 7.2-64570-4 or later, 7.1.1-42962-7 or later, 6.2.4-25556-8 or later, BSM 1.1-65374 or later, DSMUC 3.1.4-23079 or later. Synology lists no mitigation and no workaround: there is no configuration change that closes this flaw, only the update. The federal deadline is November 21, 2026, but the exploitation is already happening — treat that date as the last day, not the target day.

Patching closes the hole; it does not evict an attacker who already got in. On any device with a Hunt 1 hit, or any device still on a vulnerable build that was internet-reachable, assume the device was entered: audit the administrator accounts, audit Task Scheduler, review the DSM and nginx logs for the full window, and restore from a known-clean backup if you find persistence you cannot explain. Then remove the exposure that made this possible in the first place: take the DSM web interface off the public internet — put it behind a VPN, restrict it to your LAN or to named management addresses, and enable DSM's auto-block for repeated failed logins and two-factor authentication for administrator accounts. Verify the fix by confirming the build in Control Panel, Update & Restore after the update, and by re-running Hunt 1 over the week after patching — a burst of attempts that stops at the patch date is the exploit failing; a burst that continues is someone still trying, and a hit after the patch date is a device that did not actually update.

Sources

Keep exploring.

All articles