Rosetta Daily · Aug 13, 2026

Scanned 9 sources, 396 raw items (36-hour window), selected 19.

Critical Vulnerabilities / Actively Exploited (KEV)

Cisco ASA/FTD flaw exploited in the wild, triggers remote DoS 🔴🔥⚠️
CVE-2026-20349 (CVSS 8.6). Cisco warns a high-severity flaw in Secure Firewall ASA and FTD software — insufficient error checking when processing HTTP requests — is being exploited in the wild, letting an unauthenticated remote attacker trigger denial of service. Added to CISA's KEV catalog on August 11.
Source: https://thehackernews.com/2026/08/cisco-asa-and-ftd-flaw-exploited-in.html

Lazarus exploited a Windows zero-day to target defense firms 🔥⚠️
CVE-2026-68820 (Windows Ancillary Function Driver for WinSock use-after-free). North Korea's Lazarus Group exploited this zero-day as part of Operation Dream Job to deliver a new backdoor against defense and aerospace companies in France, Germany, Brazil, and India. Added to CISA's KEV catalog on August 11.
Source: https://www.bleepingcomputer.com/news/security/lazarus-hackers-exploited-windows-zero-day-to-target-defense-firms/

SharePoint auth-bypass flaw exploited after public PoC release 🔴⚠️
CVE-2026-55040 (CVSS 9.1). Attackers began exploiting this critical SharePoint security-feature-bypass flaw — rooted in weak authentication — after a proof-of-concept was published. Patched in Microsoft's July 2026 Patch Tuesday.
Source: https://thehackernews.com/2026/08/attackers-exploit-sharepoint.html

VMware vCenter directory-traversal flaw exploited for persistent remote access 🔴⚠️
CVE-2026-59310 (CVSS 9.8). Security firm QUIRSO reports active exploitation of this recently patched critical Broadcom vCenter flaw, which lets a network-adjacent attacker execute arbitrary code.
Source: https://thehackernews.com/2026/08/attackers-exploit-vmware-vcenter.html

Attackers target critical Adobe Commerce/Magento flaw to hijack customer accounts ⚠️
CVE-2026-71362. Exploitation attempts have been detected against this critical Adobe Commerce and Magento flaw.
Source: https://www.bleepingcomputer.com/news/security/hackers-exploit-critical-adobe-commerce-flaw-to-hijack-customer-accounts/

"ShieldBreak" zero-day PoC claims Microsoft Defender patch bypass with SYSTEM access ⚠️
Researcher Chaotic Eclipse (aka INFINITE NIGHTMARE/MSNightmare) released a PoC bypassing the patch for CVE-2026-50656 ("RoguePlanet", CVSS 7.8), potentially compromising Microsoft Defender for SYSTEM-level access.
Source: https://thehackernews.com/2026/08/shieldbreak-zero-day-poc-claims.html

Vendor Advisories

Adobe patches three CVSS 10.0 flaws in ColdFusion and Campaign Classic 🔴
The most severe, CVE-2026-48362, is an OS command injection flaw in ColdFusion that could enable arbitrary code execution.
Source: https://thehackernews.com/2026/08/adobe-patches-three-cvss-100-coldfusion.html

SAP patches maximum-severity SAP Commerce Cloud flaw 🔴
CVE-2026-58231 (CVSS 10.0). Insufficient authorization checks and input validation in the Data Hub Adapter could let an unauthenticated attacker execute arbitrary code.
Source: https://thehackernews.com/2026/08/sap-commerce-cloud-flaw-could-let.html

Web Security Research / Industry News

Malicious LiteLLM releases tied to the Trivy hack may have exposed 2,100+ organizations
Two malicious LiteLLM releases sat on PyPI for about 40 minutes in March, carrying code that harvested cloud keys, SSH keys, Kubernetes tokens, and database passwords. CloudSEK's analysis of ~434,000 stolen files maps potential exposure to more than 2,100 organizations.
Source: https://thehackernews.com/2026/08/malicious-litellm-releases-tied-to.html

"City-Forum" data-theft campaign targets Salesforce and ServiceNow portals
An ongoing campaign uses custom tools to steal data exposed to anonymous users through Salesforce Experience Cloud and ServiceNow customer portals.
Source: https://www.bleepingcomputer.com/news/security/city-forum-data-theft-attacks-target-salesforce-servicenow-portals/

737 fake Chrome VPN extensions found routing traffic through a proxy
The extensions impersonated well-known VPN/proxy services, mainly targeting Russian-speaking users, and routed traffic through SOCKS5 proxy infrastructure operated by a single provider. Combined installs exceed 75,000.
Source: https://thehackernews.com/2026/08/737-chrome-vpn-extensions-caught.html

WindRelay Android NFC relay malware paired with SpyNote RAT relays victims' card data live
The combo steals live card data and sends it to attackers in real time, used to take out loans and make purchases.
Source: https://www.bleepingcomputer.com/news/security/android-malware-combo-takes-out-loans-and-relays-victims-credit-cards/

"Plug and Pwn" attack abuses fake USB devices for Windows SYSTEM access
Researchers disclosed attacks that abuse Windows Plug and Play to trigger installation of vulnerable vendor software and gain SYSTEM privileges.
Source: https://www.bleepingcomputer.com/news/security/plug-and-pwn-attack-uses-fake-usb-devices-for-windows-system-access/

AI Security

OpenAI, Anthropic, Google API flaw let weaker models decode stronger models' hidden reasoning ⚠️
A flaw in encrypted reasoning objects used to carry hidden reasoning between API calls let a block created in one session be replayed into another; during testing, researchers recovered internal reasoning and secrets — including API keys and passwords — from session logs.
Source: https://thehackernews.com/2026/08/openai-anthropic-google-api-flaw-let.html

SpatialJB: how text distribution art becomes a "jailbreak key" for LLM guardrails ⚠️
This paper shows spatially structured token perturbations (redistributing tokens across rows/columns/diagonals) achieve near-100% attack success rate against leading LLMs, and still exceed 75% even against advanced output guardrails like the OpenAI Moderation API — outperforming prior jailbreak techniques.
Source: https://arxiv.org/abs/2601.09321

VIPER-MCP finds 106 zero-days across 39,884 MCP servers, 67 CVEs assigned so far
The first end-to-end automated vulnerability-auditing framework for Model Context Protocol servers combines static taint analysis with dynamic exploit confirmation; a large-scale scan of real-world open-source MCP repos confirmed 106 zero-days with concrete exploit traces.
Source: https://arxiv.org/abs/2605.21392

BrowseSafe: understanding and preventing prompt injection in AI browser agents ⚠️
A new benchmark of prompt injections embedded in realistic HTML payloads emphasizes injections that influence real-world agent actions rather than just text output, plus an evaluation of existing defenses and a proposed multi-layered defense strategy.
Source: https://arxiv.org/abs/2511.20597

Other

Hardware wallet maker Trezor discloses breach affecting nearly 14,000 customers
Customer data was exposed after shipping/logistics provider ShipMonk was hacked.
Source: https://www.bleepingcomputer.com/news/security/trezor-discloses-data-breach-affecting-nearly-14-000-customers/

See also: separate AI Frontier and Ransomware briefs published today (Clop's mass disclosure wave, the emerging crpx0 group, and more).