The U.S. Cybersecurity and Infrastructure Security Agency (CISA) has added a critical remote code execution (RCE) vulnerability in Microsoft's Configuration Manager, identified as CVE-2024-43468, to its catalog of known exploited vulnerabilities. This action confirms the flaw is actively exploited and mandates federal agencies to apply patches.
Vulnerability Details
Affected Product: Microsoft Configuration Manager (formerly SCCM)
Identifier: CVE-2024-43468
CVSS Score: 9.8 (Critical)
Exploitation Status: Actively Exploited
Risk & Impact
Triage: Immediate patching is required. CISA has ordered Federal Civilian Executive Branch agencies to secure systems by March 5th.
Attack Vector: A remote, unauthenticated attacker can exploit this vulnerability by sending specially crafted requests to a target server. This can lead to arbitrary code execution with the highest privileges on the server or its database.
Ease of Exploit: Proof-of-concept exploit code was made public in November 2024, increasing the likelihood of attacks.
Action Plan
Immediate Action: Apply the security updates released by Microsoft during the October 2024 Patch Tuesday. An in-console update is required for protection.
Relevant professional terms
CISA (Cybersecurity and Infrastructure Security Agency)
A U.S. federal agency responsible for cybersecurity and infrastructure protection across all levels of government, coordinating with states, and protecting against cyber threats.
Remote Code Execution (RCE)
A type of vulnerability that allows an attacker to execute arbitrary commands or code on a target machine from a remote location, often without prior authorization.
A critical pre-authentication remote code execution vulnerability, identified as CVE-2026-1731, is being actively exploited in BeyondTrust's Remote Support and Privileged Remote Access products. This flaw allows unauthenticated attackers to execute arbitrary operating system commands, and on-premise customers are urged to apply patches immediately.
Vulnerability Details
Affected Product: BeyondTrust Remote Support (versions 25.3.1 and prior) and Privileged Remote Access (versions 24.3.4 and prior).
Identifier: CVE-2026-1731.
CVSS Score: 9.9 (Critical).
Exploitation Status: Actively Exploited.
Risk & Impact
Triage: Critical. Immediate patching is required for all on-premise appliances.
Attack Vector: An unauthenticated, remote attacker can send specially crafted requests to a vulnerable appliance to execute arbitrary OS commands. No user interaction is required.
Ease of Exploit: A proof-of-concept (PoC) has been published online, lowering the bar for exploitation.
Action Plan
Immediate Action: Upgrade Remote Support to version 25.3.2 or later. Upgrade Privileged Remote Access to version 25.1.1 or later.
Workaround: No specific workarounds have been provided; immediate patching is the only recommended mitigation. SaaS instances were automatically patched by BeyondTrust on February 2, 2026.
Detection: Attackers have been observed abusing the get_portal_info function to extract the site-specific x-ns-company header. This value is then used to establish a WebSocket channel for command execution.
Relevant professional terms
Remote Code Execution (RCE)
A class of vulnerability that allows a malicious actor to execute commands of their choosing on a remote machine over a network, such as the internet.
Pre-authentication
A type of vulnerability that can be exploited by an attacker without needing to log in or provide any valid credentials. This makes it particularly dangerous as it can be exploited by any unauthenticated party.
Romania's national oil pipeline operator, Conpet S.A., has confirmed a significant data breach following a ransomware attack. The Qilin ransomware group has claimed responsibility, stating they exfiltrated sensitive company data.
Attack Overview
Attack Path: Qilin affiliates typically gain initial access via phishing emails, exploiting unpatched vulnerabilities, or using compromised credentials for services like RDP and VPNs.
Attacker:Qilin Ransomware
Impact Assessment
Data Stolen: The attackers claim to have stolen nearly 1 TB of documents, including financial records and passport scans.
Operational Impact: The company's corporate IT infrastructure was breached and its website was taken offline. However, Conpet reports that its core oil and gas transport operations and OT systems were not affected.
Strategic Takeaway
This attack on critical infrastructure highlights a frequent tactic where corporate IT networks are targeted to exfiltrate data for extortion, even when operational technology (OT) systems remain secure.
Relevant professional terms
Ransomware
A type of malicious software designed to block access to a computer system or data, often by encrypting files, until a sum of money is paid.
Double Extortion
A tactic used by ransomware groups where they not only encrypt the victim's data but also steal it, threatening to publish the sensitive information online if the ransom is not paid.
Researchers have identified numerous malicious Chrome extensions masquerading as legitimate productivity tools. These extensions are designed to steal sensitive data, including credentials, session cookies, and browsing history, impacting thousands of users.
Key TTPs
Initial Access: Users are tricked into installing extensions from the Chrome Web Store or through social engineering and phishing campaigns.
Execution: Malicious JavaScript is injected into web pages to intercept network traffic, exfiltrate data from web forms, and capture authentication tokens.
Defense Evasion: Extensions use obfuscation and may delay malicious behavior to avoid detection. Some can also block access to administrative and security pages to prevent removal.
Campaign Analysis
This threat highlights a persistent risk layer where attackers abuse the browser's permission model to conduct widespread data theft. The campaigns often involve sophisticated methods, including compromising legitimate developer accounts to push malicious updates.
Targeting & Infrastructure
Target Profile: The campaign targets general Chrome users as well as employees of enterprises using HR and ERP platforms like Workday and NetSuite.
Infrastructure: Attackers use their own command-and-control (C2) servers to exfiltrate stolen data and send commands to the malicious extensions.
Relevant Terms
Session Cookies: Small pieces of data a website stores on a user's browser to remember them and keep them logged in during a session.
Command-and-Control (C2): A server or infrastructure used by attackers to send commands to compromised systems and receive stolen data.
The North Korean Lazarus Group is targeting developers by planting malicious software packages in the npm and PyPI open-source repositories. This campaign, codenamed "graphalgo," uses fake job offers on platforms like LinkedIn to lure developers into using the compromised code, ultimately aiming to steal data and cryptocurrency.
Key TTPs
Initial Access: Social engineering via fake job offers on LinkedIn, Facebook, and Reddit.
Execution: Victims are tricked into downloading and running malicious dependencies from npm or PyPI as part of a fake coding assessment.
Defense Evasion: The campaign uses a token-based authentication mechanism for its command-and-control (C2) communication to hide its activity.
Campaign Analysis
This operation marks a strategic shift by Lazarus to weaponize the software supply chain, exploiting the inherent trust developers place in open-source repositories. The modular design allows the attackers to easily swap front-end personas while reusing the core malicious infrastructure.
Targeting & Infrastructure
Target Profile: Software developers, particularly those working with JavaScript, Python, blockchain, and cryptocurrency technologies.
Infrastructure: Malicious packages hosted on npm and PyPI, with fake company websites and GitHub repositories used for social engineering.
Relevant Terms
npm: Node Package Manager, the default package manager for the JavaScript runtime environment Node.js, used to manage and share reusable code.
PyPI: The Python Package Index, a repository of software for the Python programming language that helps users find and install software developed by the community.
Microsoft has uncovered a widespread campaign of "AI Recommendation Poisoning," where companies manipulate AI assistants through hidden instructions in features like "Summarize with AI" buttons to bias their future responses. This technique is simple to execute using freely available tools.
Key Findings
Over a 60-day period, Microsoft detected over 50 unique instances of this attack.
The campaign involves 31 companies across 14 different industries, including sensitive sectors like healthcare and finance.
Attackers use hidden prompts in URL parameters to inject persistent commands into an AI's memory, instructing it to favor a specific company or product.
The Bottom Line
This emerging threat represents a new attack surface that bypasses traditional security controls. It fundamentally undermines the trustworthiness of AI-driven decision-making tools by subtly manipulating their outputs over the long term. Leaders must now treat the integrity of AI memory and input channels as a critical security concern, as poisoned recommendations can influence strategic decisions in areas from financial investments to technology procurement without any obvious signs of compromise.
Relevant Terms
AI Recommendation Poisoning: A technique where an attacker injects hidden, unauthorized instructions into an AI assistant's memory to make it produce biased or malicious recommendations in future interactions.
Prompt Injection: The method of hijacking an AI's output by sending it hidden instructions within a seemingly harmless request or piece of content.
Wiz Research has launched the AI Cyber Model Arena, a benchmark to test the offensive capabilities of AI agents against 257 real-world security challenges. It provides a standardized method to measure how effectively AI can handle offensive security tasks.
Key Features
Real-World Challenges: Evaluates AI agents on 257 challenges across five domains, including zero-day discovery and CVE detection.
Broad Coverage: Tests capabilities in API, web, and multi-cloud security for AWS, Azure, GCP, and Kubernetes.
Deterministic Scoring: Uses programmatic scoring for objective, repeatable results without human-in-the-loop or LLM-based judging.
Use Case (The "So What?")
This benchmark offers critical insights for both offensive and defensive teams. Red teams can identify the most effective AI model-and-agent pairings for specific tasks, while blue teams can better anticipate and defend against emerging AI-driven attack vectors by understanding their capabilities.
Availability
The AI Cyber Model Arena is a research project and benchmark report from Wiz. The results are publicly available on the company's website.
Relevant Terms
Zero-day: A software vulnerability that is discovered and exploited by attackers before the vendor is aware of it or has developed a patch.
CVE (Common Vulnerabilities and Exposures): A standardized system for identifying and cataloging publicly disclosed cybersecurity vulnerabilities.
