Risky Bulletin: Brotli RCE is wrecking what's left of the internet

0:03While everyone was distracted by PHP, another RCE has been wrecking the internet. Australia's ABC wipes some staff devices, a third PHP RCE is incoming, and researchers discover a new DDoS attack vector in HTTP3. This is the Risky Bulletin, prepared by Katalin Kimpanu and read by me, Claire Aird. Today is the 26th of October, and this podcast episode is brought to you by Nebulok. In today's top story, a newly disclosed remote code execution vulnerability in Google's Brotley data compression library impacts a large number of web server and traffic management technologies. Any server that decompresses Brotley data is affected.

0:50The Brotley format became widely popular after Google open-sourced it more than a decade ago. The blast radius is just as large as last week's PHP 2 shell vulnerabilities. The Brotley bug has been confirmed to impact Node.js, Python apps, the Bunn JavaScript runtime, the Envoy proxy and Istio traffic management apps. Many other servers are expected to be impacted. Meanwhile, the PHP security team is preparing patches for a third remote code execution bug. The PHP Foundation issued a fresh security bulletin on Friday. No patch is available yet, but the mitigations that were effective against last week's two remote code execution bugs also work against this new one.

1:37The PHB Foundation says this bug is unlikely to be exploited due to it relying on an unusual configuration to be exploitable. In other news, the Australian Broadcasting Corporation wiped a small number of staff devices last week over fears they were exposed to spyware. The devices had loaded ads from an Israeli company that's known to deliver zero-click spyware. The ABC says it found no signs of a compromise and wiping the devices was a precautionary measure. The Light LLM project has linked a recent supply chain attack to the leak of a master key in May. A threat actor used the key to add an info stealer to a new Light LLM version released last week.

2:21Their master key was for the Lite LLM agent platform, an experimental tool that hosts cloud agents that interact with GitHub repositories. A hacker has gained access to the internal business systems of AI cloud platform company Genesis. The intruders entered via a learning management system and a customer relationship management platform. The company says it didn't find any evidence of the hackers pivoting to its customers' clouds. Hackers have stolen credentials from customer accounts on the Andromeda cloud AI platform. Andromeda says the intrusions were part of an industry-wide attack on AI infrastructure providers.

3:04Threat actors are compromising accounts, stealing AI API keys and renting out access to AI infrastructure on underground forums. European energy giant Engie has confirmed a security breach. The incident impacted customers of its gas supply business in France. The hackers are believed to have gotten in via an unnamed CRM platform used by business partners to sell energy contracts. Engie is notifying affected customers. The Australian government has sanctioned two North Korean companies and two individuals involved in remote IT worker schemes. Sanctions were levied on individuals Sim Hyeon-Sop and Yoo Pee-Oong and businesses Korea Meng Yong-Dae Computer Technology Company and Qingyang Information Technology Cooperation Company.

3:57All four have been sanctioned by the US since 2023. Sweden has banned government workers from using chat GPT and clawed AI assistants. Government staffers can only use approved AI assistants, such as Microsoft Co-Pilot. The ban came into effect on Monday. The Greek Parliament has passed a bill to ban kids under the age of 15 from social media. The new law will come into effect next year. Greece will become the first EU country to impose such a ban after a similar law in France was struck down by its top court. The Russian government has extended a pilot program that tracks foreigners across Moscow to the entire country.

4:42All foreign nationals entering Russia will have to install a mobile app on their phones that tracks their location. The Kremlin made the app mandatory after the Krakos City Hall terrorist attack in 2024. The U.S. Federal Communications Commission has banned the sale of two devices in the U.S. for containing foreign surveillance chips. The banned devices are a dashboard camera and a wireless video transmitter, but the FCC hasn't named the specific products. Both use chips from HiSilicon, a Huawei subsidiary. The U.S. administration has blacklisted Huawei and designated the company a threat to U.S.

5:21national security. CISA's Chief of Vulnerability Response, Lindsay Sukovnik, is leaving the agency after six years. Under her tenure, CISA rolled out the highly successful KEV program. CISA has not announced a replacement. Dutch police have arrested two suspects over a mobile phishing and NFC relay fraud campaign. The suspects posed as banks and tricked users into installing malicious apps on their smartphones. They then convinced the victims to scan their cards with the app and enter their pins. The card and pin information was relayed to other criminals who were standing by at ATMs or payment terminals and used to make fraudulent purchases or withdraw cash.

6:08Microsoft's security team has identified a cluster of malicious extensions that launch malicious behavior when users who have them installed view a malicious ad. The ads trigger browsing data collection, ad injection and shopping link modification. Microsoft has spotted at least 250 of the malicious extensions across the Microsoft Edge add-ons, Chrome Web Store and Mozilla add-ons portals. Threat Actors can crash modern HTTP3 web servers by exploiting flaws in the QPAC compression format. Malcrafted QPAC packets can uncompress into huge volumes of data, hundreds and even thousands of times the size of the original packet, exhausting the target server's memory.

6:53The BUN Framework's implementation of HTTP3 is the most vulnerable, with a packet amplification factor of almost 4 ,000. Other HTTP3 server types are also vulnerable, but not to the same degree. And finally, Meta has disabled a Muse feature that lets users share private AI conversations after reports of data leaks. The feature created public URLs for the chats that anyone could discover and scrape for sensitive information. Both ChatGPT and Grok had similar exposures last year. And that is all for this podcast edition. Today's show was brought to you by our sponsor, Nebulok. Find them at nebulok.io.

7:38Thanks for your company.

7:47you