- Browser security policies fully blocked Optimizely attack at 24 organizations and prevented data theft at another 17 out of 159 assessed, says Report URI founder Scott Helme
Recent Microsoft and US aviation disruptions, alongside a cyberattack involving Optimizely, have exposed how failures in shared technology can spread across organizations, with experts warning that backups and alternative suppliers may retain the same underlying vulnerabilities.
The incidents had different causes, but each showed how a problem in a critical dependency could affect services far beyond its starting point.
“Shared services concentrate risk. When thousands of organisations depend on one provider, a single failure or compromise at that provider becomes thousands of incidents at once,” Scott Helme, a security researcher and founder of Report URI, told Anadolu.
Microsoft failure spread despite recovery systems
Microsoft’s preliminary report puts its main disruption between Nov. 12 and 13, with Exchange Online email hardest hit and dependent features in Teams, Copilot and SharePoint also affected.
The company traced the failure to a software defect introduced in July, exposed after Teams traffic shifted to a network gateway also serving Exchange Online. Nearly 14% of that gateway’s server capacity was already unavailable because of a separate issue.
As servers stopped responding, automated systems redirected traffic to healthy ones, triggering the same defect there. At the worst point, more than 80% of servers in the affected gateway infrastructure were unresponsive.
Barriers intended to contain failures did not prevent the cascade. Some services could switch away from the faulty gateway, but Exchange Online could not do so at the required scale, Microsoft said.
Recovery required a software fix and the gradual restart of groups of servers. Microsoft said its tests missed the defect because laboratory hardware and traffic differed from production conditions.
The company described the incident as a technical failure unrelated to a cyberattack.
Cybersecurity expert Candid Wuest said customers’ recovery preparations must distinguish between preserving information and maintaining operations.
“Backups often focus on protecting data rather than maintaining the service,” he told Anadolu.
“For example, during the M365 outage, having a backup of your messages did not necessarily let you send and receive email.”
An alternative can also fail alongside the primary system if both depend on the same cloud infrastructure, identity service or network, he added.
FAA disruption reached 17 control centers
In aviation, a telecommunications problem disrupted flight schedules across the United States.
The Federal Aviation Administration imposed a nationwide ground stop on Nov. 13, with departures resuming nearly five hours later. Instability affected 17 of its 22 Air Route Traffic Control Centers and spread to several facilities handling arriving and departing aircraft.
The FAA said data flooded its network at roughly 1,000 connection attempts every four to five seconds, disrupting communications and radar feeds.
Its initial investigation identified two possible sources: a radar site in Amarillo, Texas, and a communications site in El Dorado, Arkansas, where network provider L3Harris had been deploying a new router. Isolating the locations stabilized the network.
The FAA reported approximately 12,462 flight cancellations that day, while stressing that not all were directly caused by the ground stop. Some radio frequencies and radar feeds remained available, and every aircraft landed safely, it said.
Transportation Secretary Sean Duffy said officials did not believe a cyberattack caused the disruption. L3Harris said it was supporting the investigation.
Trusted website code became an attack route
The Optimizely incident demonstrated how a shared supplier could become a route for malicious code.
Report URI identified infected website scripts associated with 178 organizations on Nov. 16, including the BBC, CNN, Zoom, HSBC, Salesforce, IKEA and Air France-KLM.
Attackers inserted malicious instructions into customer projects on Optimizely’s website experimentation platform, allowing code to reach visitors through a domain the websites already trusted.
Report URI said the malware could steal payment information and cookies, replace cryptocurrency wallet addresses and deliver exploits targeting iPhones and iPads. Exposure did not mean every visitor suffered a successful compromise.
Optimizely confirmed that an unauthorized party exploited a vulnerability to add malicious JavaScript to some customer projects. It said it fixed the vulnerability, removed the injected code and introduced additional protections.
“In the Optimizely case, attackers didn't need to breach the BBC, CNN, Zoom or HSBC individually. None of those organisations' own systems were touched,” Helme said.
He pointed to Content Security Policy, or CSP, which lets websites restrict where browsers load scripts from and send data.
“Of the 159 affected organisations whose policies we could evaluate, the CSP fully blocked the attack for 24, including Associated Bank, and stopped the data theft for another 17,” he said.
Wuest cautioned that CSP cannot necessarily stop malicious instructions already embedded in an allowed script.
For wider resilience, he urged companies to obtain evidence about suppliers’ infrastructure and subcontractors, alongside contractual commitments.
“A different provider does not necessarily mean different infrastructure; both may rely on the same cloud, identity platform or network,” he said.
Because customers often struggle to verify those dependencies, transparency and practical tests of switching to alternative services are particularly important, he added.
“The key distinction is between recovering your data and keeping your business running.”