Accueil > Accueil > Global Security Mag Online
http://www.globalsecuritymag.fr/
Articles
-
Vigilance.fr - libgit2 : multiples vulnérabilités du 20/08/2026
20 octobre, par Vigilance.frUn attaquant peut employer plusieurs vulnérabilités de libgit2, du 20/08/2026.
Voir en ligne : https://vigilance.fr/vulnerabilite/... -
Vigilance.fr - libgit2 : multiple vulnerabilities dated 20/08/2026
20 octobre, par Vigilance.frAn attacker can use several vulnerabilities of libgit2, dated 20/08/2026.
View online : https://vigilance.fr/vulnerability/... -
16-year-old Broadcom Wi-Fi flaw exposes millions of routers
20 octobre, par Valentin Jangwa, Global Security Mag — afficheA 16-year-old vulnerability in Broadcom Wi-Fi software could allow attackers to take control of millions of routers without knowing their network passwords, according to research published by security researcher Quentin Kaiser.ㅤ
The flaw affects Wi-Fi Protected Setup (WPS) software used across dozens of manufacturers and hundreds of products.ㅤ
Kaiser demonstrated unauthorized code execution on routers used by three major Belgian internet providers, including Proximus, Orange and VOO. He estimates approximately 1.56 million affected devices in Belgium alone.ㅤ
The vulnerability allows attackers within Wi-Fi range to extract credentials or execute commands on affected equipment.
ㅤ
Broadcom acknowledged knowing about the issue since 2018 but declined to assign a CVE identifier, citing the software's obsolete status. Although fixes exist for some platforms, Kaiser reported that the tested devices remained unpatched as of October 15. Older devices may require replacement.
ㅤ
Experts with BreachLock and Suzu Labs offer perspectives on the matter.
ㅤ
Seemant Sehgal, Founder & CEO, BreachLock:
ㅤ
“This is not the first case of software being classified as obsolete but still deployed in millions of devices, and it will not be the last. The gap between what a vendor is responsible for maintaining and what still handles live traffic leaves real exposure."
ㅤ
“For ISPs, now is the time to review whether the devices they ship to customers have a documented end-of-support date, because the lack of one is how 1.5 million routers end up running vulnerable firmware with no path to a fix. For home users, the practical takeaway is that routers have a shorter useful security life than most people realize, and if yours came from your ISP more than a few years ago, it is probably running software that no one is responsible for keeping secure anymore."
ㅤ
Jacob Krell, Senior Director: Secure AI Solutions & Cybersecurity, Suzu Labs:
ㅤ
“The most consequential detail in Quentin Kaiser's Broadcom Wi-Fi research is a paperwork decision. Broadcom has known since 2018 that its Wi-Fi Protected Setup (WPS) parsing code lets anyone in radio range execute commands on a router without knowing the network password. It confirmed a patch exists, then declined to assign a Common Vulnerabilities and Exposures (CVE) identifier because the software is obsolete. That software still runs on roughly 1.56 million gateways at Proximus, Orange and VOO in Belgium, and on gear from dozens of other vendors, including seven TP-Link models listed in Kaiser's disclosure, from the Archer VR600 and VR900 at Türk Telekom to the VR1600 at Australian ISPs like Internode. Obsolete is doing a lot of work in Broadcom's sentence."
ㅤ
“A CVE identifier is what turns a vulnerability into an operational fact. Without one, the flaw cannot enter the Cybersecurity and Infrastructure Security Agency's (CISA) Known Exploited Vulnerabilities (KEV) catalog, Qualys and Tenable have nothing to write a detection rule against, and no procurement checkbox pressures an internet service provider (ISP) to act. CERT/CC closed Kaiser's case the same day he filed it, because Broadcom is a CVE Numbering Authority and said no."
ㅤ
“I have yet to see a patching program with an answer for a vulnerability that has no CVE to scan for. Kaiser gave everyone 90 days, and as of October 15 every device he tested was still unpatched. Replacement, rather than patching, is the realistic fix for most of these fleets, and ISP hardware refresh cycles run on decades.” -
DHS bans DeepSeek AI products across federal civilian agencies
20 octobre, par Valentin Jangwa, Global Security Mag — afficheThe Department of Homeland Security has ordered federal civilian agencies to remove Chinese AI company DeepSeek's products from government IT systems and prohibit future purchases, according to a Federal Register notice published October 20.
ㅤ
The order also covers DeepSeek affiliates and third-party products or services developed using DeepSeek technology. DHS cited concerns that sensitive government information could be transferred to China, along with potential malware, security vulnerabilities and data collection risks.
ㅤ
Agencies must remove affected products or seek an exception. DHS also recommends enhanced network monitoring, logging and controls to detect potential data exfiltration. The order does not apply to private businesses or consumers.
ㅤ
An expert with BreachLock offers perspective on the matter.
ㅤ
Seemant Sehgal, Founder & CEO, BreachLock:
ㅤ
“The order prompts the harder internal question for private sector security leaders, which is whether the data an employee pasted into DeepSeek last month to summarize a document is now part of someone else's training set. Shadow AI usage has been running ahead of governance in most enterprises for at least the past year. The specific risk the DHS order names, which is that queries and attached data pass through infrastructure outside allied jurisdictions, applies to any employee using any foreign AI service that offers free access.”
ㅤ
“The affiliates and third-party products clause in the order is where this gets harder, because an employee does not need to use DeepSeek directly to send data to it. A commercial product built on DeepSeek technology would route the same queries through the same infrastructure without the user ever knowing which model was handling the request.” -
OpenCode breach exposes AI credentials, triggers unauthorized charges
20 octobre, par Valentin Jangwa, Global Security Mag — afficheAI coding platform OpenCode disclosed a security breach affecting 1.4% of its Console users after attackers accessed a legacy system, potentially exposing API keys for OpenAI, Anthropic and Google AI services.
ㅤ
The October 16 attack targeted OpenCode Console, which manages users' AI provider credentials. According to TechCrunch, affected users reported unauthorized AI usage charges ranging from hundreds to more than $1,000. OpenCode described the incident as a large-scale AI model distillation attack.
ㅤ
OpenCode notified affected users on October 17, advised them to revoke and replace their API keys, and contacted OpenAI, Anthropic and Google. The company has not disclosed how attackers gained access.
ㅤ
An expert with BreachLock offers perspective on the matter.
ㅤ
Seemant Sehgal, Founder & CEO, BreachLock:
ㅤ
“When a single AI developer tool holds API keys for OpenAI, Anthropic, and Google, breaking into that tool gives an attacker immediate billing access against three of the largest frontier model providers at once. That is why some users saw fraud charges reach the four figures before the breach was even disclosed.”
ㅤ
“For organizations using any AI coding platform, provider credentials need the same rotation discipline as production database passwords, with per-project keys and spending caps set at the provider level so one compromised tool does not translate into unlimited usage against every account the organization owns. Pentesting the AI tool chain itself, specifically what a compromised developer tool could reach and spend before detection, is a test worth adding to the regular rotation.”
GS Days