Find what you’re looking for
Browse everything
6,838 results in Threat Reports
CISA Adds Collabora Online RCE and Arista VeloCloud SQL Injection to KEV: What Network Security Teams Must Do Now
In October 2026, CISA added two critical vulnerabilities to its Known Exploited Vulnerabilities catalog: CVE-2026-110447 (CVSS 9.2), an OS command injection flaw in Collabora Online's Built-in CODE Server enabling remote code execution via proxy.php on Nextcloud platforms, and CVE-2026-114625 (CVSS 9.9), an SQL injection vulnerability in Arista VeloCloud Orchestrator allowing remote attackers with only read-only access to reach privileged internal functionality. Active exploitation of CVE-2026-110447 involved web shell implantation on Nextcloud servers followed by credential harvesting, while CVE-2026-114625 exploitation was traced to two IP addresses geolocated to Japan, marking the third VeloCloud vulnerability weaponized since July 2026. This incident underscores an accelerating pattern of critical infrastructure and SD-WAN orchestration platforms being targeted through high-CVSS injection vulnerabilities. With three VeloCloud flaws actively exploited within a four-month window, attackers are systematically probing network management planes—raising the stakes significantly for organizations relying on these platforms for hybrid and multi-cloud connectivity.
Operation IXIM: How an Abandoned Popcorn Time Updater Nearly Became the World's Largest Accidental Botnet
Between November 2025 and March 2026, Dutch firm Eye Security discovered that the abandoned Time4Popcorn application—a legacy fork of the popular piracy client Popcorn Time—contained a background updater (CVE-2026-30612, CVSS 9.8) that phoned home over unencrypted HTTP to unregistered domains, installed payloads at SYSTEM/root privilege without signature verification, and exposed an estimated 2.5–3 million Windows machines and up to 350,000 Macs to silent takeover. Eye registered the abandoned domains, confirmed inbound traffic from government, military, and corporate networks, and—finding no vendor to patch the software—pushed over 2 million benign disabling updates under Operation IXIM, reducing active call-home traffic by 98 percent before transferring domain custody to Shadowserver in March 2026. This incident illustrates a growing supply-chain blind spot: software that organizations never formally adopted—installed by employees on personal or corporate devices—can harbor persistent, privileged execution channels years after the original vendor disappears, creating botnet-ready infrastructure at scale without any malicious actor needing to write a single line of exploit code.
CVE-2026-111497: The Brotli Decoder Flaw That Put Node.js, Bun, and Python at Risk of Remote Code Execution
In October 2026, a critical memory-corruption vulnerability (CVE-2026-111497) was disclosed in Google's Brotli decoder library, affecting a wide range of software including Node.js, Bun, Python Brotli packages, Envoy, and Istio. The flaw, rated CVSS 4.0 score of 9.2 by Google, allows crafted Brotli-compressed data to overwrite memory boundaries, enabling potential remote code execution, crashes, or memory leaks. Imperva's Red Team demonstrated unauthenticated RCE on an unpatched Node.js 24.22.0 build. Patches were released on October 21, 2026, but each affected product must independently update its bundled copy of the decoder. No in-the-wild exploitation had been confirmed as of initial disclosure. This supply-chain vulnerability illustrates the systemic risk of widely-shared C library code embedded across language runtimes, proxies, and package ecosystems. As organizations increasingly rely on microservices and API-driven architectures, a single upstream flaw can propagate silently across hundreds of services, making runtime network policy enforcement and egress controls essential complements to patching workflows.
Inside the LiteLLM 1.107.0 Supply Chain Attack: How PRIVCOLLECT Stole Cloud and AI Credentials via Python Startup Hooks
On October 21, 2026, attackers published a malicious LiteLLM 1.107.0 release to PyPI after abusing stolen developer GitHub credentials to hijack the upstream release workflow. The compromised wheel silently planted Python startup files—cook.pth, cook.py, and sitecustomize.py—outside the application directory, causing a credential-harvesting payload (PRIVCOLLECT) to execute automatically whenever Python initialized in any affected environment, including developer workstations, CI/CD agents, containers, and AI gateway hosts. Collected credentials were AES-256-CBC encrypted, RSA-key-wrapped, and exfiltrated to an attacker-controlled domain, while a secondary Unix loader enabled dynamic remote code execution via a GitHub-hosted dead-drop pointer. This incident illustrates how a single compromised open-source dependency can silently harvest cloud credentials, AI model-provider API keys, SSH keys, and browser secrets across an entire organization's software supply chain—without the application ever being actively invoked. The growing reliance on AI gateway libraries like LiteLLM in production environments amplifies the blast radius significantly.
One Unrevoked Key, Total Pipeline Compromise: Inside the 2026 LiteLLM Supply Chain Attack
In May 2026, a LiteLLM Agent Platform master key was accidentally committed to a public GitHub repository and never revoked. On October 10, 2026, an unknown threat actor exploited this unrevoked credential to access the Agent Platform, extract a broadly scoped GitHub CLI OAuth token belonging to an organization owner, and harvest secrets from internal CI/CD workflows via malicious cache.yml files pushed to short-lived throwaway branches. On October 21, the attacker redirected the release pipeline to a lookalike repository and published the backdoored LiteLLM 1.107.0 to PyPI, Docker Hub, and GHCR—delivering a credential-stealing infostealer and persistent Python backdoor to downstream users across the AI developer ecosystem. This incident exemplifies the compounding risk of a single unrevoked secret in an AI-native supply chain: one leaked master key cascaded into organization-wide repository access, CI/CD pipeline hijack, and mass credential exfiltration targeting AI API keys, cloud credentials, and developer tool tokens. As agentic AI frameworks become critical infrastructure, secrets hygiene and pipeline integrity controls are emerging as top-tier enterprise security priorities.
php2shell Exposed: Why This PHP RCE Exploit Is Harder to Pull Off Than the Hype Suggests
In October 2024, the cybersecurity community reacted urgently to the disclosure of a PHP vulnerability nicknamed 'php2shell,' which theoretically enables remote code execution through a complex heap manipulation chain. Analysis by SANS ISC researcher Bojan revealed that reliable, weaponized exploitation is significantly harder than initial reports suggested. A working exploit requires precise memory pressure conditions, a single-worker Apache prefork configuration, knowledge of libphp.so offsets to defeat ASLR, and commands limited to 14 bytes—conditions rarely met in production environments with default 128 MB memory limits and multi-worker setups. Organizations are advised to set session.upload_progress.enabled=0 immediately as a mitigation. This incident highlights the growing challenge of accurately triaging vulnerability severity in real-world cloud and hybrid environments. As AI-generated proof-of-concept code floods public repositories, security teams face increasing noise in distinguishing theoretical exploitability from practical risk—making network-level controls and rapid configuration hardening essential layers of defense.
Eight Max-Severity ArubaOS Flaws Expose Enterprise Wireless Infrastructure to Unauthenticated RCE
In October 2026, Hewlett Packard Enterprise released emergency security patches addressing 72 vulnerabilities in ArubaOS (AOS), the operating system powering Aruba Networking enterprise wireless infrastructure including mobility conductors, access points, controllers, and gateway devices. Eight of these flaws carry maximum CVSS severity scores and enable unauthenticated remote code execution and authentication bypass across both AOS-8 and AOS-10 software versions. The vulnerabilities—spanning command injection, SQL injection, and buffer overflow vectors—require no privileges and no user interaction, making them trivially exploitable at scale across HPE's installed base of over 55,000 enterprise customers, including 90% of Fortune 500 companies. HPE confirmed no active exploitation or public proof-of-concept code existed at time of disclosure. This disclosure follows a pattern of critical HPE network infrastructure vulnerabilities in 2025–2026, including StoreOnce auth bypass flaws, hardcoded Aruba credentials, and actively exploited HPE OneView weaknesses flagged by CISA. Attackers increasingly target network infrastructure operating systems as a high-leverage initial access vector, making unpatched edge devices a primary risk surface for enterprise environments.
PHP CVE-2026-115560: How to Defend Against an Unpatched RCE Vulnerability Targeting Millions of Web Servers
In October 2026, the PHP security team disclosed CVE-2026-115560, a remote code execution vulnerability affecting PHP installations with non-default session serialization configurations. Exploitable only when session.serialize_handler is switched at runtime before session_start() with upload progress enabled, the flaw carries high severity despite its conditional nature. No patch was immediately available; mitigation required disabling session.upload_progress.enabled. This disclosure followed emergency patches for two related critical RCE flaws, CVE-2026-109618 and CVE-2026-109619 (collectively PHP2Shell), which were already seeing broad exploitation attempts across millions of exposed PHP hosts globally before patches were released. This incident underscores accelerating attacker timelines against unpatched web infrastructure. With over 5.1 million PHP hosts exposed online and exploitation of PHP2Shell beginning four days before patches shipped, organizations running PHP-based applications face shrinking windows between disclosure and active compromise, demanding proactive mitigation and network-layer controls beyond patch cycles alone.
CVE-2026-30612: How an Abandoned App's Expired Domain Put 3 Million Devices at Risk of RCE
A critical remote code execution vulnerability (CVE-2026-30612) was discovered in Time4Popcorn, an abandoned fork of the Popcorn Time BitTorrent client, affecting an estimated 3 million Windows and 350,000 macOS devices. The flaw resided in the app's auto-updater, which contacted an expired domain over unencrypted HTTP and accepted unsigned executables—enabling anyone who registered the domain to push arbitrary code with SYSTEM-level or root privileges. Researchers at Eye Security registered the abandoned update domain and 14 related domains, then used the same vulnerable mechanism to deploy remediation payloads that disabled the updater across approximately 2 million devices between November 2025 and March 2026. Connections originated not only from residential ISPs but also from corporate, government, and military networks, amplifying the potential blast radius significantly. This incident exemplifies the growing threat of abandoned-software supply-chain attacks, where unpatched updaters tied to expired domains become persistent, low-cost entry points. As organizations accumulate shadow IT and forgotten consumer applications on managed endpoints, adversaries increasingly scout lapsed domains as cheap vectors for mass exploitation—making software lifecycle governance and egress traffic enforcement critical defensive priorities.
BlackX UEFI Bootkit: How APT41 Exploits a 'Patched' Windows Flaw to Own Diplomatic Networks
In early 2026, APT41-affiliated threat actors deployed a sophisticated UEFI bootkit dubbed BlackX against diplomatic organizations in Tajikistan, Uzbekistan, and Turkmenistan. Derived from the notorious BlackLotus bootkit, BlackX exploits CVE-2022-21894 (Baton Drop) to bypass Secure Boot on fully patched Windows systems where administrators failed to apply mandatory manual revocation steps—specifically blocklisting the 'Windows Production PCA 2011' certificate in the UEFI DBX. After gaining administrator privileges through an undisclosed initial access vector, attackers installed a kernel-level driver, a RAT using Google Drive and OneDrive for C2, and the OxiDLLRunner DLL sideloader, achieving persistent, covert espionage access across high-value government networks. This incident underscores a growing APT trend of repurposing commercial crimeware tooling—bridging cybercrime and state-sponsored espionage—to maintain firmware-level persistence that survives OS reinstalls and endpoint detection. As UEFI-based implants proliferate among nation-state actors, organizations face pressure to close patch-management gaps that go beyond standard Windows Update cycles and require manual firmware-layer hardening.
Ruby's Regex Engine Had Three Bugs. GitLab Got Four Critical RCEs—Two Rated CVSS 10.0.
In October 2026, Ruby disclosed three critical bugs in Onigmo, its built-in regular expression engine, that collectively enabled four remote code execution vulnerabilities in GitLab—two rated CVSS 9.9 and two rated a perfect 10.0. GitLab compiles user-supplied regex patterns in features such as CI/CD configuration, and a specially crafted pattern could corrupt memory during compilation, granting an unauthenticated attacker full code execution on GitLab's servers. The unauthenticated flaws, CVE-2026-97229 and CVE-2026-110264, affect every GitLab version from 7.0 onward. Patched Ruby builds (4.0.9, 3.4.12, 3.3.13) and corresponding GitLab releases (19.5.1 down to 18.11.13) were shipped between September 23 and October 21, 2026. This incident highlights a systemic supply-chain risk: a single language-level regex engine flaw propagated into multiple critical application CVEs. As DevSecOps pipelines rely heavily on Ruby-based tooling and GitLab-hosted workflows, organizations must treat upstream language runtime patches with the same urgency as direct application patches, and implement inline traffic inspection to detect exploitation attempts before patches can be applied.
AI-Executed Ransomware: Inside the Phantom Framework That Automated Intrusions Across 62 Organizations
In October 2026, cybersecurity researchers at SOCRadar exposed a sophisticated AI-driven intrusion ecosystem operated by affiliates of the INC and The Gentlemen ransomware groups. The operation centered on an autonomous framework called Phantom, which used OpenCode and commercial AI models via OpenRouter to execute human-approved but AI-automated kill chains. Attackers exploited edge devices including FortiClient EMS, SimpleHelp, cPanel, and VPN gateways using multiple CVEs, then moved laterally, harvested credentials, and exfiltrated sensitive data via reverse-SOCKS tunnels before deploying ransomware. Approximately 62 organizations were targeted, with over 50% based in the United States, primarily across industrial, manufacturing, energy, healthcare, and pharmaceutical sectors. This incident signals a structural shift in the ransomware threat landscape: AI agents are now operationalizing intrusion playbooks at machine speed while keeping humans in an approver role. The use of WormGPT-MCP, Ghost C2, and AI-curated victim knowledge bases indicates that agentic AI is rapidly becoming a standard component of ransomware-as-a-service toolkits, compressing attacker timelines and raising the bar for enterprise defenders.