Key Points
- Optimizely has confirmed attackers exploited a vulnerability in its Experimentation platform to inject malicious JavaScript into 188 scripts identified across 178 organizations. BleepingComputer reports the account signup flow as the specific entry point; the full technical details have not been publicly disclosed by Optimizely.
- The malware swapped cryptocurrency wallet addresses across multiple blockchain networks including BTC, LTC, ETH, TRON, Solana, TON, BCH, and DOGE. Approximately $7,040 in observed attacker-wallet inflows were recorded as of 19:25 UTC, after funds had already been swept to exchanges at 13:03 UTC. These are on-chain observations only; card theft, session theft, and mobile exploitation losses are unquantified.
- Ledger’s website was identified as serving infected scripts during the attack window. There is no evidence in available reporting that Ledger hardware, firmware, or private keys were compromised.
- Optimizely has removed injected code and notified affected customers. Cleanup was largely complete by 21:52 UTC on November 16, with TalkTalk the last remaining infected script.
The Optimizely supply chain attack on November 16, 2026 compromised Web Experimentation scripts across 178 organizations, including Ledger, CNN, BBC, Salesforce, Zoom, GitLab, IKEA, HSBC, Equifax, Fox News, KLM, and Air France. The injected code acted as a crypto address swapper, payment card skimmer, and iOS exploit delivery mechanism, all served through a trusted CDN that affected sites had no reason to distrust.
The attack, now being tracked under the name OPTIMJACK, was identified by security researcher Scott Helme of Report URI, who found 188 infected Optimizely JavaScript files after analyzing 1,611 scripts across the platform.
How the Optimizely Supply Chain Attack Worked
Optimizely’s Web Experimentation product allows organizations to run A/B tests and personalization experiments by embedding JavaScript snippets into their websites. Because these scripts are delivered through Optimizely’s own CDN infrastructure, browsers and security tools treat them as trusted content from a known domain.
Optimizely has confirmed a vulnerability in its platform was exploited. BleepingComputer identified the account signup flow as the specific entry point, though Optimizely has not published full technical details. Optimizely’s CDN served exactly what it was given, including its own sites, which were also affected. Once inside, the attacker added a loader to each customer’s “Project JavaScript” field. The rest of the snippet remained unchanged, so it still arrived from a trusted Optimizely CDN domain.
Three loader variants were observed: a plain script tag pointing to 1688360.cc/track.js, the same URL hidden with Base64 and atob(), and a full inline copy that runs via eval() if the remote domain is blocked. When a visitor landed on an affected page, their browser loaded the Optimizely script as normal, which silently fetched and executed the attacker’s secondary payload.
Analysis by Hudson Rock and researcher Kirk (Ohey) revealed a critical resilience mechanism: the BBC and HSBC UAE copies of track.js contained a 24,392-byte Base64 fallback payload embedded directly in the script. The loader first tries to fetch the remote track.js; if that fails or times out after 2.5 seconds, it executes the embedded copy locally. This means blocking the attacker’s domain at the network level was not sufficient to stop address swapping. Removing the injected code from the Optimizely project itself was required to fully remediate the threat.
The payload was rebuilt 26 times during the day. The final variant went live at 17:57 UTC, following multiple edits within a two-hour window.
The Crypto Theft Angle: Address Swapping and Clipboard Hijacking
For crypto users, the most direct risk was address replacement. The malicious script monitored pages for cryptocurrency wallet addresses and silently swapped them across multiple blockchain networks: Bitcoin, Litecoin, Ethereum, TRON, Solana, TON, Bitcoin Cash, and Dogecoin. The script validated address checksums before substituting, so the replacement address would appear legitimate to a user who did not verify character by character. Address swapping targeted page text, forms, links, the clipboard, outgoing network requests, and even the iOS share sheet.
Critically, the attack also intercepted clipboard operations. A user who copied a wallet address from anywhere while an affected tab was open in their browser could paste the attacker’s address, even if the original address came from a completely separate source outside the compromised page.
Near wallet or payment-related content, the script also collected the page URL, any visible original addresses, JavaScript-accessible cookies, and browser screenshots.
Scott Helme tracked eight attacker wallets. Four received no funds. The attacker moved funds out at 13:03 to 13:05 UTC, hours before cleanup began. By approximately 19:25 UTC, the wallets still showed total observed inflows of roughly $7,040: 0.08298 BTC (approximately $5,380), 17.02 LTC (approximately $762), 662.9 USDT plus 184 TRX on TRON (approximately $719), 0.055 ETH (approximately $100), and 82 USDT on BSC. The 19:25 figure reflects cumulative inflows, not funds still held at that time. USDC was traced to a HitBTC deposit address.
Hudson Rock’s analysis confirmed the hardcoded destination addresses embedded in the payload. The Bitcoin address is 19g9rZcPeiz8rUvyWpTRyXXhPnjsWrc3kW, TRON is THvs6p56XkUiMkJeSc51cAP64u2RsTYqxa, EVM chains use 0xC4026b06D82b0F0B363eda0642D61CfbDc17884e, Solana uses FVtL8NMYS9PA1v3HE67XUGCtGA9JaZUvqav2fU2LrV5n, and TON, Litecoin, Bitcoin Cash, and Dogecoin each have dedicated addresses. Operator ownership of these wallets has not been attributed.
One additional detail points to a seasoned operation: the TRON wallet used in this attack has been active since March 2026, meaning the actor predates this campaign and has likely run similar attacks before.
“Observed wallet inflows do not capture potential card theft, session theft, or mobile exploitation. Those losses have no on-chain footprint.”
CryptoNewsBytes Analysis
Why Ledger Users Face Specific Risk
Ledger’s website was confirmed among the 178 organizations serving infected scripts. Cleanup on Ledger’s site was completed at approximately 20:41 UTC on November 16. The available evidence does not indicate any compromise of Ledger hardware, firmware, or private keys. The attack operated at the website layer only.
However, Ledger users who visited the site during the window and initiated any transaction should take one specific precaution: hardware wallets display the destination address on the device screen before signing. Any user who verified the full address on their Ledger device before confirming would have seen a mismatched address if swapping had occurred. Users who approved without checking the device screen may have sent funds to an attacker-controlled address.
Ledger has not issued a public statement as of this writing.
Attack Timeline
Beyond Crypto: Card Skimming and iOS Exploits
The card skimmer was not present from the start. It was added mid-attack at 13:43 UTC, capturing Luhn-valid card numbers, expiration dates, and CVV codes from forms and POST requests, sent via sendBeacon to the attacker’s server. Hosted payment iframes were out of reach.
The iOS exploit branch routed devices by OS version. iOS 13 through 17 were directed to the Coruna kit via path /A8/a.html. iOS 18 through 25 received DarkSword via /D11/b.html. iOS 26 devices were fingerprinted and routed to /C10/b.html but were not confirmed to execute a payload during testing. The router used two hosting domains: 5vobauua3ojpklx[.]icu and eil786npto9u5bq[.]icu, with telemetry sent to endpoints including /api/track, /api/gs_state, and /chain_log. Affected pages on iOS 18.6 or earlier reloaded every 60 seconds, likely to maximize exploit delivery attempts.
Kirk’s static analysis of the DarkSword kit found XOR-encoded strings using the key d7K9mP2x, a C2 base at je4ft3gn8vxudum[.]icu, a once-per-24-hour delivery gate, and a 30-minute pause after two consecutive failures. The router infrastructure and channel markers link this kit to the Caishen iOS C2 V2 ecosystem identified in earlier research by Kirk at Ohey, indicating a kit-as-infrastructure relationship though not confirming the same operator. No successful device compromises have been confirmed as of this writing.
Scott Helme’s investigation also links the same actor to an earlier wave from November 13 to 16 using scripts served through Kissmetrics, StorifyMe, and Minute Media, affecting a separate set of sites not in the Optimizely infected list. Hudson Rock’s analysis further corroborates a minute-ly loader that self-removes its own script elements three seconds after execution, a deliberate anti-forensics technique to reduce detection in browser developer tools. The same infrastructure points to the same threat actor operating across multiple script providers simultaneously.
Who Was Targeted and Why CSP Failed Most Organizations
The 178 affected organizations spanned every major industry. Banking and finance led with 32 organizations, followed by technology and SaaS (23), retail and fashion (20), streaming services (12), and news and media (10). Named victims include HSBC and Equifax in financial services, IKEA in retail, Peacock in streaming, and CNN, BBC, and Fox News in media. Hudson Rock’s direct file recovery also confirmed Zoom, GitLab, Hootsuite, Stamps.com, Tricentis, and Virgin Media O2 among those serving infected scripts. Researcher Kirk’s archived captures added Sydney Morning Herald, FAZ, Security Bank, BuzzStream, Mutual of Omaha, and Fox Sports to the documented list. These are confirmed from recovered files and archived captures, not audience estimates.
Helme also analyzed Content Security Policy enforcement across 256 organization hosts. The results were damaging: 68 organizations had no CSP at all. Of those with one, 50 had policies that allowed both the payload to run and data to be sent to the attacker. Only 9.4% of organizations, 24 of the 256 checked, had a restrictive enough script-src policy to block the payload entirely. Even allowing cdn.optimizely.com explicitly in a CSP offered no protection, since the malicious code arrived from that trusted domain. The use of 'strict-dynamic' and 'unsafe-eval' in policies let the attack through.
What Makes This Attack Different
Most crypto theft involves compromising a user’s device, phishing their seed phrase, or exploiting a vulnerability in a specific protocol. OPTIMJACK did none of those things. It hijacked the distribution infrastructure of a legitimate, trusted marketing platform and used it as a delivery vehicle.
The affected organizations did not write the malicious code. They did not configure it. Their users had no reason to distrust it. The Optimizely CDN domain is not on any blocklist. Depending on configuration, endpoint security tools may not flag a known CDN domain serving modified scripts. This is what makes supply-chain attacks against third-party JavaScript providers particularly difficult to defend against: the trust relationship is baked into every page that loads the script.
For crypto users specifically, the clipboard hijacking component is the most dangerous element. Address swapping on the page at least requires you to be interacting with a wallet interface on an affected site. Clipboard hijacking operates silently across any copy-paste action in the browser, regardless of which tab the action happens in.
What to Do If You Visited an Affected Site on November 16
Action Checklist
- Review any crypto transactions sent on November 16. Compare the destination address you intended against what actually appeared on-chain. Use a block explorer to verify.
- Always verify the full address on your hardware wallet screen before signing, not just the first and last four characters.
- Check your payment card accounts for unauthorized charges and report anything suspicious to your issuer immediately.
- Update your iPhone or iPad to the latest iOS or iPadOS version to patch vulnerabilities targeted by the Coruna and DarkSword kits.
- Rotate session tokens and passwords for accounts where sensitive credentials or sessions may have been exposed on any of the 178 affected sites.
What This Story Does Not Establish
The on-chain wallet amounts (~$7,040) represent observed attacker-wallet inflows as of 19:25 UTC and cannot be fully attributed to this campaign alone. The TRON wallet predates the campaign. Card theft, cookie theft, and mobile exploitation losses are unquantified and do not appear on-chain. No confirmed iOS device compromises have been reported. Optimizely has confirmed a vulnerability in its platform was exploited; full technical details of how the attacker gained access to customer project scripts have not been publicly disclosed. The 178 organization count covers only the 1,611 scripts Helme was able to examine. The actual scope is likely broader.
CryptoNewsBytes will update this article as Optimizely publishes further incident details or confirmed victim losses emerge.
Sources: Scott Helme / Report URI (primary investigation), Infostealers.com / Hudson Rock (payload and iOS analysis), BleepingComputer, StepSecurity (OPTIMJACK technical analysis), DeFi Prime, CyberSecurityNews

