2013 · United States · J - Information and communication · Theft of customer account data via web breach
Analysed 2026-11-06 by Hermes
Open this case in the tool →Theft of customer account data via web breach
Expertise to Execution
Evidence coverage: 2/6 stages, weight 12/60
Decisive stage: egress (clearest evidence) · Alternative: Strategy to Leadership
(1) A consumer marketplace holding more than 50 million accounts was expected, on plain state-of-practice grounds, to store credentials with an algorithm fit for password storage and to minimise the personal data it retained. (2) In practice, attackers took names, email addresses, dates of birth for some users and hashed passwords for 50 million users; the hashes were salted but stored with SHA1, a fast general-purpose hash that experts documented as an extremely poor choice for password storage, so a large percentage of the 50 million passwords was judged trivially crackable and usable for account takeover elsewhere; the company also still stored data of customers who had closed their accounts, so they too had to be notified. The entry vector was never disclosed (vendor SQL-injection / unpatched-framework hypotheses are speculation) and is not load-bearing, and detection and dwell time were never documented. (3) This is a knowing-doing gap inside the material column: bcrypt, scrypt and PBKDF2 were known specialist remedies, and LivingSocial deployed bcrypt itself within hours of the reporting - the remedy existed and could be executed, but had never been realised in the production architecture; no source shows leadership was warned and deferred, so the break is execution, not leadership (Expertise to Execution). A secondary Strategy to Leadership pattern sits in the response: the disclosure's claim that the hashes were "difficult to decode" was contradicted by expert analysis and by the company's own same-day algorithm switch - a contributing communication gap, but not decisive because the load-bearing gap is the egress one. (4) RM26 (Expertise to Execution lens): the cryptographic remedy specialists knew existed - purpose-built slow hashing (bcrypt, scrypt, PBKDF2) - was never implemented in the live password store, which stayed on SHA1 despite no technical barrier (the switch was executed within a day post-disclosure), so the stolen data was more crackable than necessary (causal role: Amplifying). RM16 (Expertise to Execution lens): data-retention and minimisation obligations were not operationalised in the company's data inventory - closed accounts' personal data was kept on live systems and was swept into the 50-million breach population instead of being purged (causal role: Amplifying).
RM26 — Data-protection / cryptographic weaknessRM16 — Data governance / minimisation gapinformationisbeautiful row: the description matches the stated victim (LivingSocial, 50 million users, part-owned by Amazon) - no data-quality mismatch. Source caveats: the case file's Naked Security link now redirects to the Sophos blog homepage (article not reachable at that URL), so the same facts were corroborated via the other contemporaneous articles cited here. The entry vector was never publicly disclosed; Imperva's SQL-injection / unpatched-Ruby-on-Rails hypotheses are vendor speculation and were not used. No regulator or forensic report on the victim-side mechanism was found (state attorneys general requested information in May 2013, but no published decision was located), which is why the mechanism rests on strong secondary reporting and the coder's classification reading. Actor unattributed - this does not lower confidence per the method spec. No related_rows in the case file. Link check: reuters.com blocks automated access (HTTP 401); the page was opened and read during research.