2020 · Afghanistan · J - Information and communication · State-sponsored cyber-espionage: repeated mail-server intrusion and data theft
Analysed 2026-11-06 by Hermes
Open this case in the tool →State-sponsored cyber-espionage: repeated mail-server intrusion and data theft
Expertise to Execution
Evidence coverage: 2/6 stages, weight 14/60
Decisive stage: detection (clearest evidence) · Alternative: Strategy to Execution
(1) Strategic expectation: as one of Afghanistan's largest telecom operators, covering all 34 provinces with over 6.5 million subscribers and holding subscriber and corporate communications, Roshan was expected to run baseline technical safeguards on its crown-jewel mail infrastructure - detection of command-and-control traffic, malware protection and egress monitoring. (2) In practice: four distinct Chinese state-sponsored clusters - Calypso APT (from at least July 2020 to September 2021), RedFoxtrot (March-May 2021) and unlinked Winnti and PlugX clusters - operated concurrently on the same Roshan mail server for over a year, beaconing to multiple C2 servers and exfiltrating at least 4 GB of data, with exfiltration spiking during the US withdrawal and Taliban takeover; Roshan detected none of it - Insikt Group found the intrusions through its own infrastructure and network-traffic tracking and notified the company before publishing, and the victim neither confirmed nor scoped the loss (exact content of the stolen data unclear, true volume unknown). (3) The classification rests on the detection stage: the knowledge that was not acted on sits in the material column (detecting C2 beaconing, blocking widely-known backdoor families like PlugX and Winnti, and watching bulk outbound transfers are routine operational remedies that require no leadership decision), and the chain broke in execution - those controls were simply not realised in operations, evidenced by the year-long dwell time and the discovery by an external party; no source documents any Roshan policy or leadership decision behind the failure, so a Strategy reading is not evidence-based, and the entry vector is unknown and not load-bearing. The egress gap (undetected bulk exfiltration) points the same way and worsened the impact; recovery was not in play and the victim's post-notification response is undocumented. (4) RM34 (Expertise to Execution lens): available detection tooling at Roshan was insufficiently configured, staffed or maintained to catch C2 traffic from four separate clusters on its central mail server, leaving a dwell time of at least 14 months that an outside firm closed instead (causal role: Obscuring). RM35 (Expertise to Execution lens): egress visibility and prevention over the carrier's network were not implemented or tuned in practice, so at least 4 GB left to multiple C2 servers undetected and the victim could not scope what was taken (causal role: Amplifying). RM28 (Expertise to Execution lens): endpoint and malware protection on the mail server did not intercept the PlugX and Winnti backdoor families - long-known, widely-tracked tools, not novel evasions - during the whole period of the intrusions (causal role: Enabling).
RM34 — Monitoring / detection / alert-handling gapRM35 — Egress / exfiltration monitoring gapRM28 — Endpoint / malware-protection gapEuRepoC row; victim, country (Afghanistan) and start year (2020, from eurepoc_extra.start_date 01.07.2020) taken from the case file and confirmed in the Insikt report. Attribution is by a private threat-intelligence vendor (technical report, IT-security-community attribution per the case file); the Chinese embassy rejected the attribution as groundless speculation (The Record). Source caveat: the source_url recorded in the dataset for the Recorded Future report (recordedfuture.com/chinese-apt-groups-target-afghan-telecommunications-firm/) now redirects to a marketing page; the cited URL is the live report page actually opened this run (redirects to /research/...), which serves the full September 2021 Insikt text; the report is also republished as a PDF by the ORKL repository. All lifecycle facts come from the vendor report and press reporting of it - no victim statement, regulator notice or governmental advisory was found in searches; the disclosure is one-sided (vendor-side). Roshan never publicly acknowledged the incident, so the response stage is recorded as undocumented rather than aligned. The RedPacket Security article was opened as a corroborating secondary source. No related rows for this victim.