← Breach analysis by the Dual Vulnerabilities framework

10 global organisations Exchange Server zero-day espionage (State-sponsored group, 2022)

2022 · O - Public administration and defence · State-sponsored espionage: data theft via chained Microsoft Exchange zero-days with webshell deployment

Analysed 2026-11-06 by Hermes

Open this case in the tool →

What happened

State-sponsored espionage: data theft via chained Microsoft Exchange zero-days with webshell deployment

Dual vulnerability type

No DV evidenced

Alternative: none defensible

Why this classification

(1) The strategic expectation is only an industry-wide one, not an organisation's: internet-facing on-premises Exchange servers were expected to be kept patched and monitored. The row describes a multi-organisation campaign - about 10 organisations worldwide, one of them a critical-infrastructure operator - and neither the case file nor any source opened names a single victim organisation, so no organisation's own policy, standard, programme, budget decision or public statement can be identified as the expectation that was not met. (2) What the sources do establish: a single activity group active from August 2022 chained CVE-2022-41040 (SSRF) and CVE-2022-41082 (PowerShell RCE), both requiring only standard authenticated access, to compromise Exchange servers, installed the China Chopper webshell for hands-on-keyboard access, and used that access for Active Directory reconnaissance and data exfiltration; Microsoft observed this in fewer than 10 organisations globally and assessed the group as state-sponsored with medium confidence (MSTIC analysis, 30 September 2022). GTSC, whose SOC found the intrusion during security monitoring and incident response for a critical-infrastructure customer, checked the exploited servers' version numbers and confirmed that the latest update had already been installed, so a previously patched ProxyShell route was impossible and the exploitation used a new 0-day; the flaws were reported to Microsoft through the Zero Day Initiative on 8-9 September 2022, patched on 8 November 2022, and the interim vendor guidance was an IIS URL-rewrite request-blocking rule that researchers showed could be bypassed. (3) Because of that combination there is no victim-side mechanism to place on either axis: the entry stage is fully explained by two unknown zero-days exploited on patched systems, so no known remedy was left unexecuted (Expertise to Execution) and no strategic artefact failed to reach operations (Strategy to Execution); the containment, detection and egress picture (webshell persistence, AD reconnaissance, exfiltration) is reported only as the attacker's activity inside a victim environment whose identity, architecture, monitoring posture and internal response are never described; and disclosure came from an IT-security company and the vendor, not from any victim, so nothing documents a notification, governance or recovery conduct of an organisation. The framework therefore has no victim situation to assess - the record confirms impact and attacker technique only - which is the decision rule's case for Unknown rather than No DV evidenced, since alignment of a victim's doctrine and execution is positively shown nowhere either. The entry vector is fully documented here but it is not load-bearing for a DV-Type; the absence of any named victim is. An Expertise to Execution reading of the sector pattern (a quarter million exposed Exchange servers, many still unpatched for ProxyShell according to researcher counts) was rejected as a group-wide pattern, not an incident-confirmed victim-side gap.

Sources

Notes

Diffuse multi-victim row: the case file carries no receiver_name (receiver_category 'Unknown;Critical infrastructure', receiver_country 'Global (region)'), so the victim can only be named descriptively and 'victim established' is false; DV-Type is Unknown per the decision rule (reporting confirms impact and attacker technique, no victim-side mechanism) and confidence is not issued. Country left null: the case file records only 'Global (region)'. Data quality: the row's description says the group stole information of 10 global organisations, while the primary source it rests on states the attacks were observed in fewer than 10 organisations globally; the row's figure was not substituted and the dataset wording is kept in the Victim field. Data quality: the critical-infrastructure operator in the row traces to GTSC's incident-response customer (GTSC does not name it and does not state its country), so no individual victim can be identified from the record. Provenance: the label ProxyNotShell for CVE-2022-41040/CVE-2022-41082 comes from researcher Kevin Beaumont (as reported by SecurityWeek and The Hacker News, both opened); the case file uses it without naming him. Dead lead: the case file's source https://msrc-blog.microsoft.com/2022/09/29/customer-guidance-for-reported-zero-day-vulnerabilities-in-microsoft-exchange-server/ no longer resolves (404 at the live target microsoft.com/en-us/msrc/blog/..., which was opened) and no Internet Archive snapshot of that post exists in the CDX index; the vendor mitigation guidance it held, and the bypass of it, are instead documented by the SecurityWeek, BleepingComputer and Microsoft pages cited above. Source caveat: the Microsoft blog notes in its 1 October 2022 update that a section on MFA as a mitigation had been included in an earlier version and removed, i.e. part of the early vendor guidance for this chain was corrected after publication. Attribution: medium-confidence technical assessment by Microsoft plus the IT-security community's China-link suspicion (webshell codepage 936, Antsword administration tool), so the actor is recorded as an unnamed state-sponsored group and no attribution beyond that is claimed. Sector (CBS SBI): the row names no organisation and its receiver categories are 'Unknown' and 'Critical infrastructure', so no single victim's economic activity applies; 'O Public administration and defence' is recorded as the least arbitrary placeholder, consistent with how other multi-target state-espionage campaign rows in this run are coded, and should be reviewed by the researchers. Analysis performed within the research boundaries: no leak site, dump or attacker communication was used; attacker-side claims are cited only as reported by the vendor and the press pages above.