← Breach analysis by the Dual Vulnerabilities framework

Uber Internal Systems Breach (Arion Kurtaj, 2022)

2022 · United States · J - Information and communication · Intrusion into internal systems via social engineering (MFA fatigue), broad lateral access to cloud and internal tooling

Analysed 2026-11-07 by Hermes

Open this case in the tool →

What happened

Intrusion into internal systems via social engineering (MFA fatigue), broad lateral access to cloud and internal tooling

Dual vulnerability type

Expertise to Execution

Evidence coverage: 3/6 stages, weight 19/60

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

Why this classification

(1) The expectation at Uber was not missing doctrine - MFA was deployed on corporate access, privileged-access management was in place, and monitoring and disclosure machinery worked - but the technical estate was expected to hold the standard identity practices known to specialists: MFA in a form fatigue cannot wear down, PAM secrets kept out of reach of ordinary accounts, least privilege around the identity backbone. (2) In practice, in a single day (September 15, 2022) an attacker using a contractor password bought from a malware-stealer feed wore the contractor down with over an hour of push-approval requests, then posed as Uber IT on WhatsApp until the contractor approved one; once inside, he read a PowerShell script on a network share that held hard-coded credentials for the Thycotic privileged-access management admin account, and from that single tool obtained tokens for AWS, G-Suite, vSphere, Duo and OneLogin - what one engineer called 'the golden ticket jackpot' - after which the breach was announced by the attacker on Slack and OpenDNS, not by Uber's controls; Uber later confirmed downloads of Slack messages and internal finance-tool data, and a London court fixed the damage at nearly $3 million. (3) This is a knowing-doing gap inside the material column: every failed control was a known, available remedy (number-matching or key-based MFA, secrets protection, privilege hygiene), not an unrecognised risk, and none of the sources shows leadership deferring a funded programme or ignoring expert warnings - so the chain broke between expertise and execution, not at leadership or at doctrine. The classification rests on the entry stage, documented by Uber's own disclosure; the containment stage (same DV-Type) is corroborated by WIRED's forensic detail; the entry vector is therefore not in doubt and, per the lifecycle rule, is load-bearing here only because the same evidence documents the failed remedy. Recovery and response were not in play as gaps. A secondary pattern - repeated credential-based intrusions after the 2014 and 2016 breaches (row 1022 covers the 2016 concealment separately) - is noted in the reporting and in Notes, but not coded, as no opened source documents a specific lesson Uber was required to implement and left unimplemented for this incident. RM22 (Expertise to Execution lens): identity experts' known countermeasure was not realised on the remote-access path - the deployed push-approval MFA stayed bypassable by fatigue plus an IT-impersonation pretext, which is how the intrusion began (causal role: Enabling). RM24 (Expertise to Execution lens): the approval procedure lacked any independent verification step, so an hour of prompts plus a WhatsApp message posing as Uber IT was enough for the contractor to authorise a stranger (causal role: Enabling). RM41 (Expertise to Execution lens): the privilege hygiene specialists prescribe was absent - reusable hard-coded administrator credentials sat in a script reachable from a contractor account, letting one login escalate to company-wide control (causal role: Amplifying). RM23 (Expertise to Execution lens): Uber's own privileged-access management tooling was turned against it - its compromise handed out tokens across cloud and identity platforms, so the blast radius equalled the PAM's reach (causal role: Amplifying).

Evidence by stage

Risk markers

Sources

Notes

Case description ('The Uber Hack's Devastation Is Just Starting to Reveal Itself; Uber Security Update') matches the stated victim Uber - it is the WIRED article of 16-09-2022 on the September 2022 intrusion plus Uber's own newsroom update; no data-quality mismatch. Employee vs contractor: the attacker, the New York Times and AP reported an employee; Uber's 19-09-2022 update and 8-K say an 'Uber EXT contractor' whose password was bought on the dark web; the classification uses Uber's version and does not depend on the distinction. Attribution: Uber initially said the attacker was affiliated with Lapsus$; a London jury in August 2023 convicted Arion Kurtaj of the Uber hack (prosecutors described it as a solo spree by a Lapsus$-associated teen; some later reporting links this circle to what became Scattered Spider) - Notes record the sequence, actor column uses the conviction. No confirmed theft of customer PII: Uber held that production systems, user accounts and sensitive user databases were not accessed; confirmed losses were internal (Slack messages, finance invoice-tool data, accessed HackerOne bug reports); trial evidence put the damage near $3 million. The case file's third source URL (uber.com/newsroom/security-update/) carries the September 15/16/19 updates of this incident; Uber posted a strikingly similar 'Security update' for a separate, minor Lapsus$ contractor episode in June 2022 - the timestamps on the opened page confirm the September incident. Related rows: 1022 is the separate 2016 breach and cover-up (own result file exists, victim naming aligned); rows 827 and 1079 carry no case files in this run's folder and no claims are made about them. SEC.gov required a declared user agent for the 8-K; all cited URLs were opened in this task. Link check: uber.com blocks automated access (HTTP 406); the page was opened and read during research. Link check: sec.gov blocks automated access (HTTP 403); the page was opened and read during research. Link check: reuters.com blocks automated access (HTTP 401); the page was opened and read during research. Unattended run: research only; no action requiring consent was attempted.