Hudson Rock Research / OPTIMJACK update
Beyond Optimizely: Malicious Scripts Traced Across Four Web Services
Optimizely has confirmed a signup vulnerability behind malicious customer-project code. Report URI reports affected Optimizely snippets associated with 178 organizations. Hudson Rock’s investigation with Kirk of Ohey documents a wider delivery picture: poisoned scripts served through Optimizely, Kissmetrics, StorifyMe and minute-ly, with archived vendor activity recorded on November 13, before the November 16 Optimizely wave. [1] [2] [3] [4]
The confirmed Optimizely access path advances the investigation. The access paths behind the other three services’ poisoned responses, the number of exposed visitors and the resulting losses remain unresolved.
These are different measures. Provider, file, organization and visitor counts cannot be added together. No verified total of compromised visitors is established by this investigation.
How attackers used Optimizely
Optimizely told BleepingComputer that an unauthorized party exploited a vulnerability in its Experimentation account signup flow and added malicious JavaScript to customer projects. Its official incident notice, posted at 01:02 UTC on November 17, confirms an Experimentation vulnerability and malicious project-code injection. [1] [2]
The recovered files show the delivery mechanism: malicious code sat inside the project JavaScript included in legitimate Optimizely snippets. Websites loading those snippets consequently brought attacker code into their visitors’ browsers. The trusted delivery hostname made the script part of the page’s existing dependency chain. [3]
Optimizely says it patched the vulnerable path, removed injected code, ended sessions, revoked tokens and reset access for affected accounts. Those account measures concern Optimizely’s affected access, not a reset of every visitor’s sessions at downstream websites. Its public incident was listed as Monitoring when reviewed. [1] [2]
The disclosure identifies the vulnerable workflow, but does not explain the bug’s code-level mechanics. It also does not explain the poisoned Kissmetrics, StorifyMe or minute-ly responses. The supported finding is unauthorized project-code publication; the reviewed evidence does not establish CDN cache poisoning or breaches of every affected brand’s own servers.
Four services carried malicious code
Kirk’s archived body checks and Hudson Rock’s recovered samples connect poisoned analytics, content-widget, experimentation and video-loader responses. Earlier responses referenced webshell[.]lol/get.js; later Optimizely files delivered 1688360[.]cc/track.js. Both paths connect to mobile-router infrastructure using channel 02111110. [3]
| Delivery provider | Recorded observation | What supports the finding | Record |
|---|---|---|---|
| Kissmetrics | November 13, 05:42 UTCsecuritybank.com | Poisoned analytics response. Kirk analyzed the historical body; Hudson Rock matched the archived URL and hash. | Archived capture |
| StorifyMe | November 13, 20:31 UTCfaz.net | Poisoned widget response. Kirk analyzed the historical body; Hudson Rock matched the archived URL and hash. | Archived capture |
| Optimizely | November 16, 04:31 UTCbbc.com | Earlier BBC response loaded webshell[.]lol/get.js. Later Optimizely assets recovered by Hudson Rock delivered track.js. | Archived capture |
| minute-ly | November 16, 16:25 UTCfoxsports.com | Poisoned video loader. The archived hash matches Hudson Rock’s directly recovered 561-byte file. | Archived capture |
These are observation times, not intrusion start times. The BBC row concerns an earlier loader variant; its later fallback-bearing file has a different hash. Representative exact resources and hashes appear in the evidence appendix.
Evidence from four poisoned delivery services does not establish four independently explained vendor-system breaches. The shared hosts, loader code and routing marker support a technical relationship. They do not independently identify one operator or prove a common way into all four services.
The November 13 records also prevent treating all linked activity as a few hours on November 16. They establish separate earlier observations, not uninterrupted malicious delivery throughout the interval.
Additional websites in the wider Optimizely assessment
Hudson Rock’s original combined report documents 21 named web properties, including BBC, Zoom, GitLab, HSBC UAE, CNN, Fox News, Fox Sports, Salesforce, FAZ and the Sydney Morning Herald. That evidence set is not an estimate of the incident’s full size. [3]
Scott Helme of Report URI supplies the broader Optimizely count: 188 poisoned snippets associated with 178 organizations. Selected additional associations from its published list are below. Hudson Rock has not independently revalidated these additional files. [4]
| Organization or brand | Reported web properties |
|---|---|
| Equifax | equifax.commy.equifax.comsignin.my.equifax.comwww.equifax.com |
| Ledger | ledger.comwww.ledger.com |
| IKEA | client.home-design.ikea.com |
| ASOS | asos.com |
| Mattel | creations.mattel.com |
| Peacock | peacocktv.comwww.peacocktv.com |
| Alaska / Hawaiian Airlines | www.alaskaair.comwww.hawaiianairlines.com |
| Typeform | typeform.comwww.typeform.com |
| Mixpanel | mixpanel.com |
| UNICEF USA | www.unicefusa.org |
For the full list, consult Report URI’s investigation. An affected subdomain does not establish exposure across an organization’s entire website, product, customer base or checkout system.
Equifax told Reuters it had stopped using Optimizely and found no evidence of compromise to its own systems. Salesforce said it was investigating unauthorized Optimizely changes and disabling scripts as a precaution. These responses illustrate why website-script exposure and backend compromise must be reported separately. [7]
The embedded fallback changes the cleanup question
Hudson Rock directly recovered eight distinct poisoned Optimizely assets and one poisoned minute-ly asset. The team also corroborated the URLs and SHA-256 values in 16 vendor-response records across 15 website captures supplied by Kirk. Historical body interpretation for the unrecovered Kissmetrics and StorifyMe files remains attributed to him. [3]
In the later BBC and HSBC UAE Optimizely files, Hudson Rock decoded the same 24,392-byte financial-theft payload embedded within the trusted response. The loader attempted to run that copy if the remote payload failed or did not arrive within roughly 2.5 seconds. Browser execution policy can still constrain this fallback. [3]
Deeper iOS stages are now documented; successful exploitation remains unproven
The earlier combined investigation linked the mobile delivery to Coruna and DarkSword. The newer findings describe what later stages were designed to do. [3]
In updated Ohey research, Kirk reports retrieving further DarkSword D11 stages, including post_exploit.js. The code describes device registration and uploads of keychain-2.db and other databases. Ohey also describes panel features intended to collect seed phrases and private keys. These findings extend the documented collection logic beyond the browser-only financial payload. [5]
Kirk also reports that an iOS 26.3-labelled request reached C10/b.html and a rce_worker_26.x.js file. However, the chain’s own log states usable_grade=DEAD weaponized=false. C10 was reachable in the later observation; that is not evidence of a functioning iOS 26 exploit. The earlier C10 article now carries this update. [5] [6]
These are Ohey’s later-stage findings. Hudson Rock has not independently reproduced those stages or demonstrated compromise of a real victim’s iPhone. A scan using an iPhone user-agent string is a delivery observation, not a Safari exploitation test. [9]
Removing the script does not reverse an earlier exposure
The directly recovered track.js replaces cryptocurrency destinations in accessible page content and browser operations, and watches for complete card details in accessible submissions. Conditional collection can include JavaScript-readable cookies and a rendered page image. The code does not show extraction of HttpOnly cookies or a bypass into isolated cross-origin payment frames. [3]
Impact depends on the page and the visitor’s actions. A substituted address could redirect a payment the visitor subsequently approves. Accessible card details could be collected if the module’s checks were met. An ordinary news-page visit does not itself establish card theft. The mobile delivery path creates a separate device-exploitation risk; the financial modules were not established in every earlier router-only response.
Report URI reported about $7,040 in wallet receipts by 19:25 UTC on November 16. Hudson Rock has not independently linked those transfers to affected visitors; this is not a confirmed incident-loss figure. [4]
Report URI found one of its previously poisoned Optimizely snippets still infected at 21:52 UTC on November 16. That was a measurement of its sample set, not a guarantee that every delivered copy had disappeared. [4]
Replacing an upstream file does not undo code already running in an open page, reverse a completed transfer or recover collected data. A cleaned origin can also coexist with cached responses. Hudson Rock’s direct minute-ly retrieval carried Cache-Control: max-age=86400; that header alone does not prove any particular browser retained it for a full day.
Website operators should preserve historical script versions, verify project code and cached copies, review collection-destination requests, and determine which sensitive pages actually loaded the affected assets. Public evidence does not yet establish a campaign-wide total of stolen cards, compromised devices or losses.
The next questions concern access and outcomes
The unresolved questions are how the poisoned responses reached Kissmetrics, StorifyMe and minute-ly; whether a common operator controlled the linked delivery; which pages and visitors actually executed each payload; and what data or funds were successfully taken. The Optimizely signup disclosure answers one access question. It does not answer those others.
This follow-up incorporates Hudson Rock’s original investigation, Kirk’s subsequent analysis, Optimizely’s disclosure and independently attributed reporting. Amirul Cyber’s early OPTIMJACK research remains an important starting reference. [8]
Sample records and investigation indicators
The four records below preserve the distinction between a recovered file and corroboration of a historical archive record. Hashes identify the recorded bytes, not every version ever served at a URL. All dates below are in 2026 and all times are UTC.
Four provider records with exact resources, hashes and provenance
Kissmetrics
securitybank.com · November 13, 05:42:03 UTC · 213 bytes
Served resourcei.kissmetrics.com/i.js
SHA-256c154f3b644cd51f2446483c7ffd2ae4f178c4b0cb54846cfb5b83215367349cf
Kirk’s historical body analysis; archive URL and hash corroborated by Hudson Rock. View archived record.
StorifyMe
faz.net · November 13, 20:31:21 UTC · 16,569 bytes
Served resourcestorifyme.xyz/widgets/faz-5746000/widget/custom/8746?account=faz-5746000
SHA-25694d174abe1e3b12b000f2bc8cbd8e9344290f4b5a8661c8d700cdc5675675fba
Kirk’s historical body analysis; archive URL and hash corroborated by Hudson Rock. View archived record.
Optimizely
about.gitlab.com · November 16, 16:17:40 UTC · 354,471 bytes
Served resourcecdn.optimizely.com/js/5113954737848320.js
SHA-256a87a2eaea9cfb74aae7283d15fbdb499b5a8ebc3fab4c453df7d62e503996240
Hudson Rock recovered the file directly; its SHA-256 matches the historical record. View archived record.
minute-ly
foxsports.com · November 16, 16:25:44 UTC · 561 bytes
Served resourceapv-static.minute-ly.com/resources/v2/loader.js?v=3
SHA-2560f2f2a27d19112e5800541daecc91eb6c77e49f9b43ed23c890b876631978533
Hudson Rock recovered identical bytes with and without the query parameter; counted as one asset. View archived record.
Direct retrieval: November 16, 21:29 UTC. The response carried Cache-Control: max-age=86400.
Financial payload hashes and additional archive context
Directly recovered track.js, 29,882 bytes:f7b84b9a94f749f8ced3f0112d0a66894e8bbc3b4aa5592021d1b8976a9bad40
Decoded fallback, 24,392 bytes, identical in the BBC and HSBC UAE files:c2a34c579e4c2e8802752c93b3a9cca0230610f3e4dd6688f4272459c7c7c162
Earlier BBC archive, November 16, 04:31:40 UTC:45e42f5fa7520f6e7c048587f025549b31f0f24be8be058c7037d1da37e37b67
Later BBC file directly recovered by Hudson Rock, November 16, 21:00 UTC:305d0b00a7bafae9f240dbaf5bfdbb19648e38e509e8992112fcb7866c2e37dc
These BBC hashes represent different observed versions. The earlier observation in the main table must not be assigned the later file’s fallback capability.
| Indicator | Evidence and role |
|---|---|
1688360[.]cc/track.js | Financial-theft payload recovered directly. |
1688360[.]cc/card.php1688360[.]cc/hit.php | Collection destinations in the recovered code; a code reference alone does not prove data arrived. |
webshell[.]lol/get.js | Shared loader referenced in the recovered minute-ly file and Kirk’s historical body analysis. |
5vobauua3ojpklx[.]icu/boot.jseil786npto9u5bq[.]icu/boot.js | Mobile loader hosts referenced in the recovered payload and recorded requests. |
l7bcsnwc7nzr3uo[.]icu/api/iptj | Telemetry request destination recorded in the supplied mobile-directed captures. |
je4ft3gn8vxudum[.]icu | Router C2 reference reported in Kirk’s analysis. |
ios-hack[.]icu/api/ip-sync/sync | Attempted request in the supplied capture; that record has no completed response. |
data-channel-code="02111110"window._k7 / window.__k7b | Code pivots for the routing channel and financial payload. Match in context. |
The original report preserves the full wallet list, further payload and stage hashes, and reported-only domains with their qualifications. Read the complete original IOC and methodology appendix. This update does not turn unverified indicators into confirmed infrastructure.
Source record
- Optimizely: Web Experimentation Security IncidentOfficial notice posted November 17, 2026, 01:02 UTC; listed as Monitoring when reviewed.
- BleepingComputer: Optimizely’s statement identifying the signup vulnerabilityLawrence Abrams; updated reporting reviewed November 17, 2026.
- Hudson Rock: original investigation, recovered hashes and archived response evidenceNovember 16, 2026. Includes Kirk’s supplied verification and the complete original IOC appendix.
- Report URI: affected Optimizely snippets and organization listScott Helme; November 16, 2026. Source of the 1,611 / 188 / 178 measurements and the additional website associations.
- Ohey: DarkSword analysis, including the script-vendor delivery and later stagesKirk; current findings reviewed November 17, 2026.
- Ohey: C10 research and its November 16 updatePublished November 5; relevant update dated November 16, 2026, 22:30 UTC.
- Reuters: responses from Salesforce and EquifaxNovember 17, 2026. Company responses distinguish third-party script exposure from an established backend breach.
- Amirul Cyber: early OPTIMJACK investigationNovember 16, 2026. Early payload analysis and the OPTIMJACK name.
- urlscan: browser and custom user-agent limitationsThe service uses headless Chrome; a supplied Safari user-agent string does not turn a capture into an iPhone exploitation test.