← Breach analysis by the Dual Vulnerabilities framework

Apollo contact database theft (Unattributed, 2018)

2018 · United States · J - Information and communication · Data theft via exposed database

Analysed 2026-11-06 by Hermes

Open this case in the tool →

What happened

Data theft via exposed database

Dual vulnerability type

Expertise to Execution

Evidence coverage: 5/6 stages, weight 26/60

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

Why this classification

(1) Apollo's business is holding and exposing to its customers a central store of some 200 million contacts, 10 million companies and 9 billion data points, including client-imported Salesforce data; the strategic expectation is that the crown-jewel database itself is authenticated, isolated from the open internet, segregated by sensitivity, and monitored for exposure - expectations sharpened in the GDPR-era for the personal data of 100+ million people who never dealt with Apollo. (2) In practice a July 2018 system upgrade left the prospect database unprotected - Wired describes the trove as readily available online for anyone to access - and an unidentified third party stole the bulk of it plus client-imported data (7 million plus internal opportunity records, almost a million for one client); Apollo did not notice the weeks-long exposure itself (the external researcher Vinny Troia found it in mid-August 2018), notified only its roughly 530 customers in late September, declined to say whether authorities were informed or what data was taken, and left the 100+ million affected individuals to be notified by HaveIBeenPwned. (3) The remedies that would have prevented or shrunk the loss - hardening and validating the upgraded environment, keeping the database authenticated and out of reach, scanning one's own estate for exposure - sit squarely in known technical expertise, and no source shows leadership deciding, deferring or funding anything here: expertise existed and the remedy was available, but it was not realised in operations or architecture, which is the Expertise to Execution gap. The classification rests on the entry stage (clearest evidence); the containment stage - client-confidential data mixed into the same reachable pool - points the same way and is why the blast radius covered ~530 clients. The entry vector itself is not further specified than the post-upgrade exposure; the egress stage (unoperationalised privacy/minimisation duties for the hoarded data) and the response stage (customer-only, scope-free disclosure decided at CEO level) show secondary Strategy to Execution and Strategy to Leadership patterns but rest on inference and do not carry the type. Recovery is not in play: theft-only incident, no disruption. RM20 (Expertise to Execution lens): the July 2018 system upgrade changed the security posture of the estate - the victim's own letter dates the discovery weeks after system upgrades in July - and the changing environment was not hardened, tested or validated against exposure before it was exploited (causal role: Enabling). RM27 (Expertise to Execution lens): the central prospect database stood internet-reachable and unprotected, a classic insecurely-exposed database that any internet scan could find, which is precisely how it was found (causal role: Enabling). RM34 (Expertise to Execution lens): Apollo's own monitoring never registered weeks of public exposure of its flagship asset; detection came from an outsider's routine scan, so the activity went unnoticed until someone else looked (causal role: Obscuring). RM39 (Expertise to Execution lens): even at disclosure the scope could not be stated - the company would not say what data was taken or how many customers were affected, and the 100+ million affected individuals were left uninformed by Apollo, notification being carried out instead by HaveIBeenPwned (causal role: Constraining).

Evidence by stage

Risk markers

Sources

Notes

informationisbeautiful row: the one-line description matches the stated victim Apollo (sales-intelligence startup, formerly ZenProspect, YC-backed, US). The case file's source_url carries a malformed trailing ' /'; the article itself opens correctly at the Wired URL cited. Timing: incident dated 23 July 2018 (Security Affairs), external notification mid-August 2018, customer notice late September 2018, press disclosure 1-5 October 2018; Year = 2018. Reporting frames differ slightly: TechCrunch/SC Media relay the victim letter's account of theft after July system upgrades, while Wired/Privacy International describe a freely web-accessible database found by scanning; both readings support the same exposed-store mechanism and no attribution of the thief exists. Whether regulators were informed was never confirmed by the company (TechCrunch); the CTO later told Wired it reported to law enforcement. GDPR exposure was raised only as a possibility by press (no regulator action found), so the egress and response stages are recorded with inference source type. No related_rows in the case file; later incidents sometimes reported at this victim were not investigated and are not relied on here.