🎯 AI
===================

Executive summary: Urban VPN Proxy, a Chrome extension with over 6 million users, was observed harvesting AI chat data across multiple platforms. The extension injects platform-specific executor scripts, overrides core browser network APIs, and forwards captured conversations to Urban VPN infrastructure.

Technical details:
β€’ The extension deploys dedicated executor scripts (examples: chatgpt.js, claude.js, gemini.js) when targeted AI platform pages load.
β€’ Injected code wraps and overrides fetch and XMLHttpRequest so all request and response payloads for the page flow through the extension first.
β€’ Extracted fields include user prompts, model responses, conversation IDs, timestamps, session metadata, and the specific AI platform/model used.
β€’ Inter-script messaging uses window.postMessage with an identifier PANELOS_MESSAGE to pass parsed data to the extension content script.
β€’ The content script forwards packaged, compressed data to the background service worker, which transmits to endpoints such as analytics.urban-vpn.com and stats.urban-vpn.com.

Analysis:
β€’ The approach is highly invasive: overriding fetch/XMLHttpRequest captures both outgoing prompts and incoming model outputs before rendering, exposing full conversation context.
β€’ Harvesting is independent of VPN functionality and enabled by hardcoded flags with no user-visible opt-out, increasing exposure risk for users who installed the extension for privacy reasons.

Detection guidance:
β€’ Monitor outbound connections to analytics.urban-vpn.com and stats.urban-vpn.com from browser processes.
β€’ Inspect loaded extension scripts for executor filenames and for patterns overriding fetch/XMLHttpRequest and using window.postMessage with PANELOS_MESSAGE.

Limitations:
β€’ Public reporting indicates the extension targeted ten AI platforms; specific historical timeline details were not fully enumerated in the source.
β€’ No CVE identifiers or named threat actor attribution were provided in the disclosed findings.

References / Tags:
chatgpt.js, claude.js, PANELOS_MESSAGE, analytics.urban-vpn.com

πŸ”Ή ai #privacy #browser_extension #data_exfiltration

πŸ”— Source: https://www.koi.ai/blog/urban-vpn-browser-extension-ai-conversations-data-collection

8 Million Users' AI Conversations Sold for Profit by "Privacy" Extensions

Privacy browser extensions misled users and sold 8 million AI chat logs, exposing sensitive conversations for profit without consent.

πŸ“’ Abus de Microsoft Power Automate : exfiltration de donnΓ©es et persistance via des flux lΓ©gitimes
πŸ“ Selon Trend Micro, des cybercriminels et act...
πŸ“– cyberveille : https://cyberveille.ch/posts/2025-09-08-abus-de-microsoft-power-automate-exfiltration-de-donnees-et-persistance-via-des-flux-legitimes/
🌐 source : https://www.trendmicro.com/vinfo/us/security/news/threat-landscape/complexity-and-visibility-gaps-in-power-automate
#Data_Exfiltration #Living_off_the_Land #Cyberveille
πŸ“’ Analyse des tactiques d'attaque sur SharePoint Online dans Microsoft 365
πŸ“ Cet article publiΓ© par Guardz analyse une **mΓ©thodologie d'attaque** ciblant **SharePoint Online** dans les environnements *...
πŸ“– cyberveille : https://cyberveille.ch/posts/2025-08-07-analyse-des-tactiques-d-attaque-sur-sharepoint-online-dans-microsoft-365/
🌐 source : https://guardz.com/blog/adversary-tactics-and-exploitation-paths-in-sharepoint-online/
#Cloud_Security #Data_Exfiltration #Cyberveille
Ransomware Attacks Focus on Data Exfiltration Over Encryption - RedPacket Security

Ransomware actors are largely eschewing encryption, focusing instead on exfiltrating data. This trend is highlighted by a report from ReliaQuest, indicating

RedPacket Security
Chemical Facilities Warned of Possible Data Exfiltration Following CISA Breach - RedPacket Security

The US Cybersecurity and Infrastructure Security Agency (CISA) has revealed its Chemical Security Assessment Tool (CSAT) was breached by a malicious actor,

RedPacket Security
Russian Coldriver Hackers Deploy Malware to Target Western Officials - RedPacket Security

Russian threat group Coldriver has expanded its targeting of Western officials with the use of malware to steal sensitive data, Google’s Threat Analysis Group

RedPacket Security