← Breach analysis by the Dual Vulnerabilities framework

Portuguese banks (30+ institutions) banking-trojan campaign Operation Magalenha (Brazilian cybercriminals, 2023)

2023 · Portugal · K - Financial and insurance activities · Phishing campaign deploying banking-trojan backdoors, credential and PII theft

Analysed 2026-11-06 by Hermes

Open this case in the tool →

What happened

Phishing campaign deploying banking-trojan backdoors, credential and PII theft

Dual vulnerability type

Expertise to Execution

Evidence coverage: 2/6 stages, weight 13/60

Decisive stage: entry (clearest evidence) · Alternative: Strategy to Execution

Why this classification

(1) The expectation at the victim side is specialist-level, not doctrinal: banks and other regulated financial institutions are expected to protect their digital channels and customer accounts against banking trojans, and the remedies here are available technical knowledge - e-mail and endpoint protection against script-bearing phishing, and threat intelligence on the Maxtrilha banking-trojan family, publicly documented since 2021 including earlier campaigns against Portuguese banking users. (2) In practice, the Operation Magalenha infection chain ran across the user base of over 30 Portuguese financial institutions: EDP- and tax-authority-themed phishing delivered obfuscated VB scripts which SentinelOne describes as a "simple, yet effective technique for evading static detection mechanisms" with relatively low VirusTotal detection ratios, executed a loader and installed two PeepingTitle backdoors that captured window titles and screenshots and stole online-banking and government-portal credentials and personal data; the campaign persisted from 2021 to a peak in May 2023 and was exposed by an external vendor, not by the institutions themselves. (3) That is a knowing-doing gap inside the material column: experts knew the remedy and the threat, yet the controls were not realised in operations (break at execution), and no source documents a strategic artefact, budget or leadership decision failing at any named institution, so the doctrinal column cannot be implicated; the entry vector is load-bearing here only in this specific sense - the documented failure of known countermeasures to block it - not because phishing entry per se implies a DV-Type. The classification rests on the entry stage (clearest evidence, primary vendor report, load-bearing); the detection stage (multi-year undetected campaign, externally disclosed) points the same way but rests on inference and is contributing only; recovery is not in play (no disruption) and response/disclosure conduct by the banks is undocumented. A condition no marker covers is described rather than coded: the compromise occurred on the endpoints of the institutions' online-banking users, so the failure sits in a distributed user base the banks' own controls only indirectly reach. (4) RM28 (Expertise to Execution lens): endpoint and e-mail protection across the targeted institutions' reach did not block a simple, known-pattern VBS loader chain for years - protection could not be effectively deployed or tuned against the campaign's delivery path, evidenced by the scripts' low detection ratios and repeated successful infections (causal role: Enabling). RM34 (Expertise to Execution lens): monitoring, detection and threat-intelligence use at the targeted institutions were insufficiently configured or maintained to catch a banking-trojan campaign against their own user base that had been publicly documented since 2021, so the campaign went undetected internally for its whole 2021-2023 duration and time-to-detect was measured in years until an external vendor disclosed it (causal role: Obscuring).

Evidence by stage

Risk markers

Sources

Notes

Diffuse multi-victim campaign: the EuRepoC row lists 24 receiver entries (with duplicates Caixadirecta/Novobanco/Citibanamex repeated and one 'Not available'), all Portugal; all named entities are banks or bank-like finance institutions except IGCP (state debt agency), so sector is taken as K for the majority. SentinelOne documents that the intrusions hit the endpoints of the institutions' online-banking users (EDP / tax-authority themed lures), not the institutions' own networks; no per-bank corporate intrusion, statement, notification or regulatory action was found in English or Portuguese reporting, so all institutional-level origin/break readings are made at the level of the campaign as a whole, and the detection-stage gap is recorded as inference. Disclosure was by the IT-security company (EuRepoC source_disclosure), not by any victim. Attribution to a Brazilian criminal group rests on SentinelOne's technical report (attribution_type: technical report); no law-enforcement follow-up naming this campaign was found (a 2023 PJ operation against bank-phishing, Vera Cruz II, and a 2026 MP alert describe different schemes and are not cited). Related row 792 (ShinyHunters claim on Santander customer data) could not be compared: row-792.json is not present in the cases-run500 folder and no result-792.json exists in the output directory, so no claim about that row is made. The victim list is kept verbatim from the case file including duplicates and the 'Not available' entry. Country from eurepoc_extra.receiver_country, Year from start_date 01.01.2023 (the underlying activity nexus dates to 2021; 2023 is the row's start year and the peak of attacks). Link check: sentinelone.com blocks automated access (HTTP 429); the page was opened and read during research.