P.K. SHARMA

Cyber security intelligence, AI governance, practitioner analysis

SonicWall's September SMA1000 fix is the newest build its 6 October advisory lists as affected

SonicWall's 1 September fix builds for two exploited SMA1000 flaws are the newest builds its 6 October advisory lists as affected by CVE-2026-102255, a 10.0 flaw with no known exploitation. Earlier 2026 SMA1000 flaws were already exploited when fixed, and this SSRF's reach gets one sentence.

By · LinkedIn · · 27 min read

Update, 14:47 BST on 12 October (afternoon): Previdian's own page shows 135 attempts from 2 addresses on 1 sensor

This briefing has been updated twice since it was written on 7 October. A third fact sizes what the exploitation reports mean, and it comes from the page that the 9 October update already cited. Nothing about the flaw, the affected builds or the fixed builds has changed.

1. The scale of what was seen. Previdian's page for the flaw, which says it was updated on 12 October, now shows 135 exploitation attempts observed by its sensors, from 2 unique attacker addresses in 1 country, seen by 1 sensor. Its daily chart shows activity on two days, 24 and 111 events, and zero on the other days it plots. The page gives the first observation as 9 October and the last as 10 October. So Previdian's own report is one honeypot sensor, two source addresses and two days (derived from the page), and Defused's separate decoys have not published a count that I found. That is evidence that someone is trying the flaw, and it is not evidence of how many appliances were reached. Previdian has still not said whether any attempt would have succeeded.

2. A date that does not match. The Hacker News repeats the 135 attempts from two addresses on 12 October, and says the first signs were detected by Defused and Previdian on 8 October. Previdian's own page says 9 October for its first observation, and the 12 October morning update above has Defused at 8 October. I follow the primary: Previdian 9 October, Defused 8 October as quoted by others. Previdian's last observation is 10 October, so on its page there is nothing seen on 11 or 12 October, which a reader should not take to mean the attempts stopped (one sensor, one vantage).

3. Nothing else has moved. I re-read SonicWall's advisory in a browser at about 14:47 BST: First Published and Last Updated are both still 6 October, and it still says there is currently no evidence that any of the vulnerabilities in the release are being exploited in the wild. CISA's catalogue is still version 2026.10.08 and does not list CVE-2026-102255, 6 days after the advisory (derived).

What to do differently: nothing in the order of actions above changes, and the reason for urgency is not the size of the attack. Two addresses are not the story. The story is that the flaw is being tried, a fix has existed since 6 October, and the advisory has not told you. Patch to 12.4.3-03670 or 12.5.0-03082 or higher, review logs from 6 October onward for requests you cannot explain, and do not read a small count on a honeypot as a reason to wait.

What changed after the earlier 12 October update. Sources: Previdian's page (updated 12 October), The Hacker News (12 October), SonicWall advisory SNWLID-2026-0017 and CISA's catalogue (version 2026.10.08), read on 12 October 2026

ItemRead earlier on 12 OctoberRead in the afternoon of 12 October
Attempts seen by PrevidianReported, count not stated in this briefing135 from 2 unique addresses in 1 country, 1 sensor, activity on two days (24 and 111 events)
First and last observed (Previdian)First 9 October at 08:49 UTCFirst 9 October, last 10 October (its page)
First reported by any firm8 October (Defused, as quoted by others)Unchanged. The Hacker News says both firms saw it on 8 October; Previdian's page says 9 October
SonicWall's advisoryLast updated 6 October; no evidence of exploitationUnchanged at about 14:47 BST
CISA catalogueVersion 2026.10.08; not listedVersion 2026.10.08; not listed

Update, 11:16 BST on 12 October: a second honeypot operator reports attempts from 8 October, and SonicWall's advisory still has not changed

This briefing was written on 7 October and updated on 9 October. Since then one more fact has appeared that moves the earliest reported date, and two statements were re-read and have not moved. The flaw, the affected builds and the fixed builds are as above.

1. The earliest reported sighting is a day earlier than the 9 October update says. BleepingComputer quotes Defused, a threat intelligence firm, as saying it has been observing unauthenticated server-side request forgery traffic against its SonicWall decoys "since yesterday" in an article dated 9 October. SecurityWeek reports on 12 October that Defused's first attempts reached its decoys on 8 October, 2 days after the advisory's date of 6 October (derived from the dates). The 9 October update gave Previdian's first sighting, 9 October at 08:49 UTC, which was 3 days after (derived). Two honeypot operators now report the same kind of traffic within 2 to 3 days of the advisory. I read Previdian's own page on 9 October, but Defused's statement only as quoted by BleepingComputer and SecurityWeek: I did not find its own post. The requests are described in detail in the reports, and I print none. Honeypots show what is aimed at honeypots, so neither report says how many real appliances were reached or whether any attempt succeeded, which Previdian says it has not established.

2. SonicWall's advisory is unchanged. I re-read SNWLID-2026-0017 in a browser at about 11:16 BST on 12 October. It still shows First Published and Last Updated as 6 October and still says there is currently no evidence that any of the vulnerabilities in this release are being exploited in the wild. It is now 6 days since that date (derived), and two outside firms report attempts in that period. The sentence was a statement about 6 October, and this is the second update in which the briefing has had to say so.

3. The catalogue and the coordinator have not moved. CISA's catalogue is still version 2026.10.08 (1,739 entries) and does not list CVE-2026-102255, 6 days after the advisory (derived). The CVE record was last updated on 7 October and the CISA coordinator's reading is still exploitation none. A flaw can be attacked for days before a catalogue lists it. The two earlier 2026 SMA1000 advisories that reported exploitation reached the catalogue within a day of an advisory that itself reported exploitation; this advisory does not.

What to do differently: nothing changes in the order of actions above, only the urgency. Treat the flaw as one that is being attempted now. Confirm each gateway is on 12.4.3-03670 or 12.5.0-03082 or higher, review its logs from 6 October onward for requests you cannot explain, and do not wait for SonicWall's advisory or the catalogue to say the word before acting: the NCSC's 5 day and Cyber Essentials' 14 day clocks run from the fix's release on 6 October.

What changed after the 9 October update. Sources: BleepingComputer (9 October), SecurityWeek (12 October), SonicWall advisory SNWLID-2026-0017 and CISA's catalogue (version 2026.10.08), read on 12 October 2026

ItemRead on 9 OctoberRead on 12 October
Earliest reported attempt9 October at 08:49 UTC (Previdian), 3 days after the advisory (derived)8 October (Defused, as quoted by BleepingComputer and SecurityWeek), 2 days after the advisory (derived); Previdian's date unchanged
SonicWall's advisoryLast updated 6 October; no evidence of exploitationUnchanged, 6 days after its date (derived)
CISA catalogueVersion 2026.10.08; not listed, 3 days after the advisoryVersion 2026.10.08; not listed, 6 days after the advisory (derived)
Whether any attempt succeededNot established (Previdian)Not established in anything read; no victim has been named

Update, 18:10 BST on 9 October: exploitation attempts are now reported, and SonicWall's advisory has not changed

At about 18:10 BST on 9 October I read three things that were not true when this briefing was written on 7 October, and re-read two that were. None changes the flaw or the builds. All of them change one claim above: that nothing was known about exploitation.

1. Exploitation attempts are reported. Previdian's page for the flaw shows its honeypot sensors first recorded exploitation attempts against CVE-2026-102255 on 9 October at 08:49 UTC, 3 days after the advisory's date of 6 October (derived from the dates), and marks the confidence as confirmed. BleepingComputer reported it the same day, quoting Previdian's founder, and says the activity is consistent with exploitation attempts but that Previdian has not established whether any attempt would have compromised a system. The report describes the requests in detail; I print none. Previdian sells exposure and threat data, and a honeypot sees what is aimed at the honeypot, so it says nothing about how many real appliances were reached.

2. SonicWall's advisory still says the opposite. SNWLID-2026-0017 shows First Published and Last Updated both as 6 October, and still reads "There is currently no evidence any of the vulnerabilities addressed in this release are being exploited in the wild." That sentence was true of what SonicWall knew on 6 October. It is now a statement that a reader needs to date, which is the point of the section below on the date of "no evidence of exploitation".

3. The catalogue and the coordinator have not moved. CISA's catalogue is still version 2026.10.08 (released 8 October at 20:09 UTC, 1,739 entries) and does not list CVE-2026-102255. The CVE record was last updated on 7 October, and CISA's coordinator still reads exploitation as none (7 October, 15:50 UTC). Both earlier SMA1000 advisories in this briefing's history reached the catalogue within a day of the advisory; this one has not reached it in 3 days (derived). Absence from the catalogue is a statement about the catalogue, not about the flaw.

4. The population. BleepingComputer says Shadowserver now tracks more than 400 SMA1000 appliances exposed online and that it is not known how many are honeypots or already patched, which is consistent with the 412 devices, 23 of them in the UK, counted on 6 October later in this briefing. I did not re-run that count.

What to do differently: treat the flaw as one that may be attacked now, because the builds above are the only fix and an appliance on 12.4.3-03526, 12.5.0-02952 or older is affected. Check that each gateway is on 12.4.3-03670 or 12.5.0-03082 or higher, check its administration and logging for requests you did not expect since 6 October, and keep the NCSC's 5 day and Cyber Essentials' 14 day clocks, which run from the fix's release on 6 October, not from the day exploitation was reported.

What changed after this briefing was written on 7 October. Sources: Previdian's page, BleepingComputer, SonicWall advisory SNWLID-2026-0017, CISA's catalogue (version 2026.10.08) and the CVE record, read on 9 October 2026

ItemWritten on 7 OctoberRead on 9 October
ExploitationSonicWall: no evidence of exploitation; CISA coordinator: nonePrevidian: attempts first seen 9 October at 08:49 UTC, confidence confirmed; SonicWall's advisory unchanged since 6 October; coordinator's reading unchanged since 7 October
CISA catalogueVersion 2026.10.04, which predates the advisory; not listedVersion 2026.10.08; not listed, 3 days after the advisory
Exposed appliances412 identified as SMA 1000 on 6 October, 23 in the UK (Shadowserver)More than 400, per BleepingComputer's reading of Shadowserver; honeypots and patched appliances not separated
Affected and fixed buildsAffected 12.4.3-03526 and 12.5.0-02952 and older; fixed 12.4.3-03670 and 12.5.0-03082 or higherUnchanged on the advisory read on 9 October

The 1 September fix build is the newest build on the 6 October affected list

SonicWall's advisory SNWLID-2026-0016, published on 1 September 2026, gave 12.4.3-03526 and 12.5.0-02952 as the platform hotfix builds that fix two SMA1000 flaws it said were being exploited. Its advisory SNWLID-2026-0017, published on 6 October, lists the builds affected by a new flaw, CVE-2026-102255, as 12.4.3-03526 and older and 12.5.0-02952 and older. The build SonicWall told customers to install 35 days earlier (derived) is now the newest build on the affected list. An appliance that did what the September advisory asked is still affected, and the builds that fix it are 12.4.3-03670 and 12.5.0-03082 or higher.

The same thing happened between July and September. The 14 July advisory, SNWLID-2026-0008, fixed in 12.4.3-03453 and 12.5.0-02835, and the September advisory lists exactly those two as its newest affected builds. Of the 221 advisories on SonicWall's PSIRT list when we read it, eight carry a CVSS of 10.0, and three of the eight are these SMA1000 advisories, published on 14 July, 1 September and 6 October, 84 days from first to last (derived). BleepingComputer reported the October advisory on 7 October and called the flaw a maximum-severity server-side request forgery (SSRF) in SMA1000 gateways. We read SonicWall's advisory, the CVE and NVD records and CISA's catalogue ourselves. Where BleepingComputer and those sources differ, this briefing follows the sources and says so.

What that does not establish. It does not say the new flaw is a way round the September fix. The advisory links the two nowhere. What we can check is that the component (the Appliance Work Place interface), the heading (pre-authentication SSRF via unintended forward-proxy), the pair of weakness classes and the score vector are the same in both advisories, and that is a similarity of shape, not a statement of cause. It does not say any appliance has been attacked through this flaw: SonicWall says it has no evidence of exploitation, on 6 October. It does not say where the affected range starts, because the advisory says "and older versions" with no floor. And a 10.0 is a score someone assigned, which the next section takes apart.

What "maximum severity" is a score of, and who gave it

BleepingComputer calls CVE-2026-102255 a maximum-severity flaw. The number is 10.0. It appears in three records but has two authors, SonicWall in its advisory and CISA-ADP in the entry that NVD repeats, and the CVE record that SonicWall itself filed does not contain it.

Where the 10.0 for CVE-2026-102255 appears and who wrote it. Sources: SonicWall PSIRT advisory, CVE Services API and NVD API, read on 7 October 2026.

RecordWhat it carriesWhat it leaves out
SonicWall advisory SNWLID-2026-0017 (the vendor)CVSS 10.0, vector AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H labelled as version 3.0; impact listed as critical on the PSIRT listHow the score was reached, and how many appliances are reachable or attacked
CVE record, SonicWall as the CNAThe description, the affected builds, CWE-918 and CWE-441Any score: the CNA container has none
CVE record, CISA-ADP containerCVSS 3.1 base score 10.0 from the same eight metric values, and an SSVC reading of exploitation none, automatable yes, technical impact total, timestamped 7 October 15:50 UTCThe basis for the SSVC reading, which is a coordinator's view of the public record at that moment
NVD recordStatus Awaiting Analysis; the only score is the CISA-ADP 10.0, marked SecondaryA score of NVD's own

The scores agree because they come from the same choices: network access, low complexity, no privileges, no user interaction, changed scope, and high impact on confidentiality, integrity and availability. We recomputed the CVSS 3.1 base score from those eight choices and got 10.0 (derived). A score of that kind says what an attacker could do if the description is right and the flaw is reachable. It does not say how reachable the interface is on your appliance, how many are exposed, or whether anyone is attacking. And scorers differ: for the stored cross-site scripting flaw in the same advisory, CVE-2026-102258, SonicWall's vector gives 5.5 and the CISA-ADP vector gives 6.1.

Method, not accusation. SonicWall wrote the advisory, assigned the CVE and chose the score, and it also sells the appliance and the fix. Volexity, whose July report is used below, sells incident response and wrote from its own client's case. Neither fact makes a statement wrong. Both mean the figures are theirs, and a reader should be able to see whose.

What the advisory says the flaw reaches, and what it leaves out

SSRF is a class of flaw in which a server is made to send a request it should not. What it means in practice depends on what the server can reach and what answers. SonicWall's only description of reach in its advisory is one sentence: a remote unauthenticated attacker "could potentially" direct the appliance to issue requests on their behalf and "reach internal functionality and perform unauthorized operations". Its heading for the flaw is "Pre-authentication SSRF via unintended forward-proxy", and the weakness list adds CWE-441, an unintended proxy or intermediary, which is also called a confused deputy.

What SonicWall's advisory of 6 October states and does not state about CVE-2026-102255. Source: SNWLID-2026-0017, read in full on 7 October 2026.

TopicStatedNot stated
WhereThe Appliance Work Place interface of SMA1000 models 6210, 7210 and 8200v. Not SMA 100, not SSL-VPN on SonicWall firewallsWhether a particular interface configuration is needed, or whether virtual and hardware units differ
WhoA remote unauthenticated attacker; network access, low complexity, no user interactionAny precondition beyond reaching the interface
ReachInternal functionality and unauthorized operations, as a possibilityWhich functions or data, or whether code execution follows from this flaw alone
Affected12.4.3-03526 and older, 12.5.0-02952 and olderThe oldest affected build
Fixed12.4.3-03670 and higher, 12.5.0-03082 and higher; workaround: noneAny interim control short of the hotfix
ExploitationNo evidence of any of the vulnerabilities being exploited in the wildWhat was checked, by whom, and for how long
IndicatorsNone: the Comments section of the advisory is emptyAny log pattern, file or behaviour to look for

The advisory covers four CVEs, not one, and the other three are all post-authentication. SonicWall describes an operating system command injection scored 7.8 (CVE-2026-102256), a path traversal through archive extraction in the Appliance Management Console scored 7.2 (CVE-2026-102257), and stored cross-site scripting in the same console scored 5.5 (CVE-2026-102258). It says the first two can lead to code execution. For the command injection its text says a remote authenticated attacker acting as administrator, while its vector reads local access (AV:L) and low privileges (PR:L). The advisory does not reconcile the two.

The shape is not new. In both earlier pairs that SonicWall said were exploited, an unauthenticated SSRF sat beside an administrator-level command execution flaw in the same advisory, and CISA lists both halves of each pair. Volexity's analysis of the July pair, which we read, shows the two used as a chain, and its investigators found that the SSRF gave an unauthenticated route to services meant to be reachable only from the appliance itself. BleepingComputer says the September pair was also chained; we did not find a primary source for that. The October advisory does not say its flaws chain, and we have found no source that says they have been. The pairing is back, which is an inference about shape and not evidence of use.

Four SMA1000 advisories in 2026, and the two that reported exploitation were listed within a day

SonicWall's four SMA1000 advisories of 2026 against the CISA catalogue. Sources: SonicWall PSIRT advisories and CISA KEV version 2026.10.04, read on 7 October 2026. Days from advisory to listing are derived.

Advisory and fix buildsSonicWall's statement that dayCISA KEV (version 2026.10.04)
SNWLID-2026-0003, 8 April. Top score 7.2. Fixed in 12.4.3-03387 and 12.5.0-02624No evidence of exploitationNone of its four CVEs listed
SNWLID-2026-0008, 14 July. CVE-2026-15409 (SSRF, 10.0) and CVE-2026-15410. Fixed in 12.4.3-03453 and 12.5.0-02835Multiple cases indicating active exploitationBoth added 14 July, 0 days after; ransomware use Known; due 17 July
SNWLID-2026-0016, 1 September. CVE-2026-83548 (SSRF, 10.0) and CVE-2026-83549. Fixed in 12.4.3-03526 and 12.5.0-02952A case indicating active exploitationBoth added 2 September, 1 day after; ransomware use Unknown; due 5 September
SNWLID-2026-0017, 6 October. CVE-2026-102255 (SSRF, 10.0) and three more. Fixed in 12.4.3-03670 and 12.5.0-03082No evidence of exploitationNot listed; the version is dated 4 October, two days before the advisory
SonicWall PSIRT advisories SNWLID-2026-0008, 0016 and 0017, CISA KEV version 2026.10.04, the CVE Services API and Volexity's 17 July report. Gaps are derived.

The April advisory's fix builds, 12.4.3-03387 and 12.5.0-02624, are among the builds the July advisory lists as affected, although 12.4.3-03434 and 12.5.0-02800 are the newest affected builds it names. So the match between one advisory's fix and the next advisory's newest affected build holds for two of the three consecutive pairs, July to September and September to October.

Intervals, all derived: 14 July to 1 September is 49 days, 1 September to 6 October is 35, and 14 July to 6 October is 84. Volexity, which helped SonicWall's July investigation, says the earliest sign of compromise it saw was on 22 June, 22 days before the advisory. All four 2026 SMA1000 entries in the catalogue carry CISA's forensic triage flag set to Yes, which none of the 15 older SonicWall entries do.

BleepingComputer's catalogue figures check out. It says CISA has added 19 SonicWall vulnerabilities to its exploited list over four years and that 13 have been abused in ransomware attacks. From the catalogue JSON itself we count 19 entries with SonicWall as the vendor and 13 with known ransomware campaign use, so the figures agree. Two details differ. The earliest SonicWall entry is dated 3 November 2021, which makes the span 4 years and 11 months, not four years (derived). And six of the 19 are SMA1000 entries, all added since 24 January 2025, four of them this year (derived). Its account of the July pair also holds: both are listed as known ransomware campaign use. For the September pair the catalogue says Unknown, and BleepingComputer does not claim otherwise.

Two earlier briefings on this site cover the catalogue side. Cisco gave customers a compromise hunting command 47 days before the same advisory said the flaw was exploited tabulates five management plane flaws that reached the catalogue, including the September SonicWall pair, each with a three day federal deadline. CISA added 34 flaws to the exploited catalogue in September, and 26 came with a three day deadline explains why that deadline is now the usual one.

"No evidence of exploitation" is a statement with a date

SonicWall's sentence, dated 6 October, says there is currently no evidence that any of the vulnerabilities in this release are being exploited in the wild. It states a position on one day. It does not say how SonicWall looked, or what it can see on appliances it does not manage. Four things in the record limit what that sentence can carry.

The 2026 pattern is attack first, fix second. In July the advisory itself reported multiple cases of active exploitation and Volexity put the earliest compromise 22 days before it. In September the advisory reported a case. For those two there was no interval from fix to attack to measure, because the fix was the first public notice of attacks already under way. If October differs, it differs in how the flaw came to light: the advisory credits a researcher at Anthropic for CVE-2026-102255 and CVE-2026-102256, while the July and September advisories credit SonicWall's own staff, with Volexity named in July for helping the investigation. A flaw reported from outside was found by a researcher, which says nothing about whether anyone hostile found it as well.

A fix is also a description of the flaw. The NCSC's update guidance says that once a vulnerability is fixed, attackers will often study it and try to write exploits, which sets up a race between those updating and those attacking. The CISA coordinator's reading in the CVE record is automatable yes and technical impact total, with exploitation none at 15:50 UTC on 7 October. How long the race takes here is not something any source states.

The catalogue's silence carries no information yet. The latest CISA catalogue we could read, version 2026.10.04, was released on 4 October, two days before the advisory, so its not listing CVE-2026-102255 cannot be read as a finding. The NCSC's active exploitation guidance also notes that the catalogue reports only exploitation that has already happened.

The exposure figure counts devices, not vulnerable devices. BleepingComputer says Shadowserver tracks over 400 internet-exposed SMA1000 appliances, although some may have been patched. On Shadowserver's dashboard the series for devices identified as SonicWall SMA 1000 totals 412 on 6 October, the latest day, and ranges from 407 to 511 over the seven days from 30 September (derived by summing the six continent series). Of the 412, 140 were in Europe and 23 in the United Kingdom. Shadowserver describes the underlying device identification report as a device population report in which "no assessment is made on the vulnerability state of the device". A device that answers like an SMA1000 may or may not be patched, and the series we read carries no build numbers.

There is a longer tail to this kind of gateway. An earlier briefing on Microsoft's weaponisation figure records that a different SonicWall VPN flaw, CVE-2024-40766, was added to the catalogue on 9 September 2024, a year before the ransomware surge Microsoft's report describes. We checked that date in the catalogue.

What this means for UK organisations

Who runs these. BleepingComputer says SMA1000 gateways are used by government agencies, managed service providers and many large corporations to give VPN access to internal applications. It does not say in which countries. Shadowserver's dashboard identifies 23 SMA 1000 devices in the United Kingdom on 6 October, within a range of 13 to 23 across the week. That is a count of devices, not of organisations or of vulnerable units. A managed service provider may run one for many customers, and a customer may not know which build it is on. We found no NCSC publication naming SonicWall or this CVE in the NCSC's RSS feed (its 20 most recent items, read on 7 October), which says nothing about what the NCSC may have said elsewhere.

The patching clocks that apply to a hotfix released on 6 October 2026. Sources: Cyber Essentials Requirements for IT Infrastructure v3.3 (April 2026), NCSC vulnerability management guidance version 2.1. Dates are derived.

RuleWhat it requiresDate for a 6 October hotfix
Cyber Essentials v3.3, Security Update ManagementUpdates within 14 days of release where the vendor calls the vulnerabilities critical or high risk, or the CVSS v3 score is 7 or above; one critical or high fix in a bundle brings the whole update inside 14 days20 October 2026
NCSC, update by defaultInternet-facing services and software: install on a test environment or backup first, then test and roll out, completed within 5 days11 October 2026, a Sunday
NCSC, responding to active exploitationCompressed timelines of under 24, 48 or 72 hours plus a cyber incident response investigation, set out for flaws on the CISA catalogueDoes not apply as written: CVE-2026-102255 is not on the version read

Whether a given SMA1000 is in scope of a Cyber Essentials assessment is a decision for the organisation and its certification body, and we have not assumed it is. If it is, SonicWall's own description (impact critical, score 10.0) puts the hotfix inside the 14 day rule, and the NCSC's timescale for internet-facing software is shorter than the scheme's. Cyber Essentials also asks that the administrative interface of a firewall, or of a network device with firewall functionality, is not reachable from the internet unless there is a documented business need and it is protected by multi-factor authentication or a short allow list. SonicWall places two of the four CVEs in the Appliance Management Console. The SSRF is in the Work Place interface, which users reach by design, and SonicWall lists no workaround, so limiting the console does not remove the exposure that matters most here.

The NCSC's alert of 27 August on internet-exposed systems and edge devices asks organisations outside operational technology for the basics this advisory tests: an accurate inventory of internet-facing systems, an understanding of each edge device's function and data flows, prompt vendor updates, and monitoring for unexpected configuration changes or outbound connections. It also asks every organisation to register for the free Early Warning service. Early Warning reports vulnerabilities and open ports from information feeds and does not scan your network itself, and we have not seen that it covers this CVE. On suppliers, the NCSC's active exploitation guidance says critical suppliers and managed service providers should be contractually liable to mitigate exploited vulnerabilities in internet-accessible systems quickly. For this CVE nobody yet says exploitation, so that trigger may not have fired. Ask anyway.

What to do, in order

Take this with you

Defender actions, in the order worth doing

  • List every SMA1000 you run or that a supplier runs for you: hardware 6210 and 7210, virtual 8200v, disaster recovery and test units. Record the exact build string of each and who is responsible for patching it.
  • Compare each build with the advisory. Affected is 12.4.3-03526 and older and 12.5.0-02952 and older. Fixed is 12.4.3-03670 or higher and 12.5.0-03082 or higher. An appliance on the 1 September fix build is affected.
  • Before patching, follow SonicWall's own hotfix article: export a configuration backup, snapshot a virtual appliance from its host, check the download against the published checksum, expect a reboot that drops connected users, and have console access ready. That article is dated 2020, so also read the technote that comes with the hotfix.
  • Apply the platform hotfix for your firmware line. SonicWall's article says hotfixes are cumulative and that a hotfix for one firmware cannot be applied to another. Aim for the NCSC's five days, 11 October, and treat Cyber Essentials' 14 days, 20 October, as the limit.
  • While you wait for the change window, restrict the Appliance Management Console to a management network or a short allow list, as the NCSC and Cyber Essentials advise. This does not mitigate the new SSRF flaw, which is in the interface users reach, and SonicWall lists no workaround.
  • Review the appliance's logs and its outbound connections for the period you ran affected builds. This advisory publishes no indicators, so look for the appliance connecting to internal addresses or services it does not normally use, and for unexpected configuration changes, as the NCSC edge device alert advises.
  • Confirm that someone did the compromise review SonicWall asked for in July, where the advisory and Volexity's report list indicators, and in September, where SonicWall asked customers to contact its support. SonicWall's own advice for a positive finding is to re-image or redeploy, change all user and administrator passwords and reset TOTP tokens, which tells you a hotfix alone is not treated as remediation.
  • Ask your managed service provider in writing which SMA1000 builds it runs for you, when each was patched and who watches for the next advisory.
  • Re-read the CISA catalogue for CVE-2026-102255 before you close the change. If it appears, the NCSC's compressed timelines apply and CISA's entries for this product have carried a forensic triage requirement since July.
  • Register the addresses of the appliances with the NCSC's free Early Warning service if you have not already.

What is not established, and what I could not read

  • Whether any appliance has been attacked through CVE-2026-102255. SonicWall said no evidence on 6 October, and no source we read says otherwise.
  • What internal functionality the flaw reaches. The advisory gives one sentence.
  • Whether the flaw is related to September's. Only the shape matches.
  • The hotfix release time and download checksums. The mysonicwall.com download page needs an account and was not opened, so a release on Tuesday 6 October rests on the advisory date and BleepingComputer.
  • How many appliances are affected, as against merely identifiable. Shadowserver counts devices and shows no build.
  • NVD's own analysis and score. The record was awaiting analysis.
  • Whether CISA will list the flaw. The catalogue version read predates the advisory.

The question

SonicWall has now put out three SMA1000 advisories with 10.0 flaws in 84 days, and in the second and third the previous fix build was the newest build on the affected list. "Patched in September" is a friendly label. A hotfix is a name and an advisory is a date, and the only control is the build string on each gateway.

So which build is each of your SMA1000 gateways running at this minute, and who in your organisation, or at your managed service provider, compares the next advisory with it?

Key facts

Sources

  1. PrimaryAdvisory SNWLID-2026-0017, first published 6 October 2026, read in full in a browser at about 17:47 BST on 7 October 2026. The primary source: four CVEs, affected and fixed builds, scores, the no-evidence statement, the empty Comments section, credits. SonicWall wrote the score and sells the product.SonicWall PSIRTaccessed 2026-10-07
  2. PrimaryAdvisory SNWLID-2026-0016, first published 1 September 2026, read in full: the 12.4.3-03526 and 12.5.0-02952 fix builds, the exploitation statement, the recommended actions.SonicWall PSIRTaccessed 2026-10-07
  3. PrimaryAdvisory SNWLID-2026-0008, first published 14 July 2026, read in full: the 12.4.3-03453 and 12.5.0-02835 fix builds, the exploitation statement, the indicator and recommended action sections, credits.SonicWall PSIRTaccessed 2026-10-07
  4. PrimaryAdvisory SNWLID-2026-0003, first published 8 April 2026, read in full: 12.4.3-03387 and 12.5.0-02624 fix builds and the no-evidence statement.SonicWall PSIRTaccessed 2026-10-07
  5. PrimaryThe PSIRT vulnerability list, read through the page's own public API at about 17:48 BST on 7 October 2026: 221 advisories, publication timestamps, the eight advisories carrying a CVSS of 10.0 and the four SMA1000 advisories of 2026.SonicWall PSIRTaccessed 2026-10-07
  6. PrimaryCVE record for CVE-2026-102255 read through the CVE Services API: SonicWall as CNA with no score, reserved 28 September, declared public 7 October 13:01 UTC and published 13:07 UTC, and the CISA-ADP container with the 3.1 vector, score 10.0 and the SSVC reading.CVE Programaccessed 2026-10-07
  7. PrimaryCVE record for CVE-2026-102258, read with the records for CVE-2026-102256 and CVE-2026-102257: the CISA-ADP scores 6.1, 7.8 and 7.2 set beside SonicWall's 5.5, 7.8 and 7.2.CVE Programaccessed 2026-10-07
  8. PrimaryNVD record for CVE-2026-102255 read at 17:46 BST on 7 October 2026: status Awaiting Analysis, the CISA-ADP 3.1 score marked Secondary and no score of NVD's own.NIST NVDaccessed 2026-10-07
  9. PrimaryKnown Exploited Vulnerabilities catalogue, version 2026.10.04, released 4 October 2026 at 18:52 UTC, 1,734 entries, fetched at 17:46 BST on 7 October 2026. Source of the SonicWall counts, the dates added, the ransomware flags, the due dates and the forensic triage field.CISAaccessed 2026-10-07
  10. PrimaryProxying to Compromise: SonicWall Secure Mobile Access 0-day Exploitation, 17 July 2026, read in full. Used for the earliest sign of compromise on 22 June and the post-exploitation findings, at defender level only. Volexity sells incident response and wrote from its own client's case.Volexityaccessed 2026-10-07
  11. PrimaryDashboard time series of devices identified as SonicWall SMA 1000, 30 September to 6 October 2026, read through the dashboard's own embedded chart data, grouped by continent and filtered to the United Kingdom.The Shadowserver Foundationaccessed 2026-10-07
  12. PrimaryDevice Identification Report description: a device population report with no assessment of vulnerability state, based on publicly accessible responses.The Shadowserver Foundationaccessed 2026-10-07
  13. PrimaryKnowledge base article How to apply or rollback Hotfix on SMA1000 Device, marked updated 22 October 2020 and read on 7 October 2026: configuration backup, snapshot, checksum, reboot, cumulative hotfixes, firmware line.SonicWallaccessed 2026-10-07
  14. PrimaryCyber Essentials: Requirements for IT Infrastructure v3.3, April 2026, read in full with pypdf: the Security Update Management requirement (14 days) and the firewall administrative interface requirement.NCSCaccessed 2026-10-07
  15. PrimaryVulnerability management, 1. Put in place a policy to update by default, version 2.1, reviewed 1 May 2026: the five day timescale for internet-facing software, install on a test environment or backup first, the race after a fix, and check for compromise before updating when exploited.NCSCaccessed 2026-10-07
  16. PrimaryVulnerability management, 2. Responding to active exploitation of vulnerabilities, version 2.1, reviewed 1 May 2026: the compressed timelines table, the limits of the CISA catalogue and supplier contract wording.NCSCaccessed 2026-10-07
  17. PrimaryDisruptive cyber activity highlights risk from internet-exposed systems and edge devices, published 27 August 2026: advice for organisations outside operational technology and the Early Warning recommendation.NCSCaccessed 2026-10-07
  18. PrimaryEarly Warning service page: free alerts including vulnerability and open port alerts, built from information feeds without active scanning by the NCSC.NCSCaccessed 2026-10-07
  19. PrimaryProtect your management interfaces, a blog post published 22 March 2017: expose management interfaces to dedicated management networks or at least limit inbound addresses. Older guidance, cited for the principle.NCSCaccessed 2026-10-07
  20. PrimaryThe NCSC's site-wide RSS feed, 20 most recent items, read on 7 October 2026 to check for any item naming SonicWall: none found.NCSCaccessed 2026-10-07
  21. Reported bySonicWall warns of max severity SSRF flaw in SMA1000 gateways, 7 October 2026, the news pointer, read in a browser in my own tab. Used for the Shadowserver figure, the 19 and 13 KEV counts, the users of SMA1000 and the earlier CVEs, each of which was checked against a primary.BleepingComputeraccessed 2026-10-07
  22. Reported bySonicWall SMA1000 flaws exploited as zero-days to push custom malware, 20 July 2026, read to check the 'exploited for weeks' claim and to find Volexity's report. The September chaining claim comes from a further BleepingComputer article that was not read.BleepingComputeraccessed 2026-10-07
  23. Reported byCVE-2026-102255 exploitation observed, first seen in honeypot sensors on 9 Oct 2026 at 08:49 UTC, confidence confirmed, not in CISA KEV; read 9 Oct 2026Previdianaccessed 2026-10-09
  24. Reported byMax severity SonicWall SMA1000 flaw now exploited in attacks, 9 Oct 2026: Previdian's observation, SonicWall's unchanged advisory, Shadowserver's count of more than 400 exposed appliancesBleepingComputeraccessed 2026-10-09
  25. Reported byCritical SonicWall Vulnerability Exploited Days After Disclosure, 12 Oct 2026: Defused's first attempts on its SMA decoys on 8 October, Previdian's report, SonicWall's disclosure of 6 OctoberSecurityWeekaccessed 2026-10-12
  26. Reported byCritical SonicWall SMA 1000 Flaw Targeted in 135 Exploitation Attempts After Patch, 12 Oct 2026: Previdian's 135 attempts from two addresses, first signs dated 8 October (not matched by Previdian's own page), SonicWall's unchanged advisoryThe Hacker Newsaccessed 2026-10-12

Share this briefing

Know someone who owns this problem? Send it to them.

Related briefings

The briefing, in your inbox

Practitioner analysis of cyber and AI security news. No vendor noise.

How often

Every new briefing in one email, at 7am, or at 7am, 12:30pm and 6pm. Nothing is sent when nothing is new. Unsubscribe any time.