Click any tag below to further narrow down your results
Links
A new attack, dubbed "Operation WrtHug," has compromised around 50,000 outdated ASUS WRT routers, primarily in Taiwan and Southeast Asia. Researchers suspect the campaign is linked to China, leveraging multiple known vulnerabilities to facilitate stealthy espionage.
This article introduces the Depthfirst platform, which improves application security by validating vulnerabilities before flagging them. It reduces false positives and offers actionable fix suggestions to streamline security efforts for development teams.
A new attack called TEE.fail compromises the security of Trusted Execution Environments (TEEs) from Nvidia, AMD, and Intel. It utilizes a simple hardware method that, once executed, renders these TEEs untrustworthy, even if the operating system kernel is compromised. This raises significant concerns about the security claims made by chipmakers regarding their TEEs.
Researchers at HiddenLayer found a flaw in the guardrails of popular AI models like GPT-5.1 and Claude. The EchoGram attack uses specific words to trick these safety systems, allowing harmful requests to bypass defenses or causing harmless requests to be flagged as dangerous.
Sweet Security offers a comprehensive solution for cloud defense, leveraging AI to identify and prioritize vulnerabilities. It provides real-time visibility and rapid response to threats, helping organizations secure their environments without frequent scans. The platform also simplifies compliance and governance processes.
This article discusses the security vulnerabilities associated with GitHub Actions, highlighting issues like secrets management failures, insufficient permission management, and dependency pinning failures. It emphasizes the importance of understanding these risks to protect CI/CD workflows from potential attacks.
This article analyzes the vulnerabilities of the Model Context Protocol (MCP) used in coding copilot applications. It identifies critical attack vectors such as resource theft, conversation hijacking, and covert tool invocation, highlighting the need for stronger security measures. Three proof-of-concept examples illustrate these risks in action.
This GitHub repository offers intentionally vulnerable Model Context Protocol (MCP) servers for security research and training. Each server includes detailed instructions for running it and demonstrating various vulnerabilities, such as code execution and data exposure. Users should only operate these servers in a controlled lab environment.
Ivanti disclosed two critical vulnerabilities affecting its Endpoint Manager Mobile, which have already compromised several major organizations, including the Netherlands’ government and the European Commission. Researchers warn that attacks are spreading, with nearly 1,300 instances still exposed online. The vulnerabilities allow attackers to execute code remotely, raising concerns over ongoing exploitation.
OpenAI warns that its upcoming AI models may pose a "high" cybersecurity risk due to their enhanced capabilities. The company reports that these models could enable more people to execute cyberattacks, especially with their ability to operate autonomously for longer periods. OpenAI is increasing its efforts to address these threats through collaboration and new tools.
SquareX's research reveals a concealed MCP API in the Comet browser that allows extensions to execute commands and access user devices without permission. This breach of security principles raises concerns about potential exploitation and the lack of transparency surrounding embedded extensions.
OpenSSL has released updates to address 12 vulnerabilities, including a critical remote code execution flaw. Discovered by Aisle, the issues mainly involve memory safety and parsing errors that could lead to denial of service or exploitation. The most severe flaws affect versions 3.0 to 3.6, particularly in CMS and PKCS#12 handling.
The article reveals how Claude Cowork is vulnerable to file exfiltration attacks due to unresolved flaws in its code execution environment. Attackers can exploit prompt injection to upload sensitive user files to their accounts without any human approval. The risks are heightened by the tool's integration with various data sources, making it essential for users to remain cautious.
Reflectiz offers a solution that continuously monitors and manages web threats like tracking pixels and malicious scripts. It helps organizations identify vulnerabilities in their websites and implement security measures effectively. The service operates remotely, simplifying the integration process.
Check Point Research identified critical vulnerabilities in Microsoft Teams that allow attackers to manipulate messages, spoof notifications, and impersonate users. Four specific types of attacks were detailed, highlighting the potential for business email compromise and identity fraud. Microsoft has issued fixes for these issues, but concerns remain about security.
A security researcher has criticized Apple's macOS bug bounty program for significantly lowering payouts for certain vulnerabilities. Despite increasing rewards for high-profile exploits, many macOS categories now offer much smaller financial incentives, which could discourage researchers from reporting flaws.
This article explores vulnerabilities in various Object Relational Mappers (ORMs), focusing on how improper filtering can expose sensitive data. It highlights specific cases in Beego and Prisma ORMs and discusses exploitation methods, including time-based attacks. The authors also provide tools for detecting these vulnerabilities.
In 2025, an AI system identified four previously unknown security issues in OpenSSL, three of which were disclosed and fixed by the system. The findings highlight the potential of AI in proactively discovering vulnerabilities in critical infrastructure.
Hackers are exploiting misconfigured web applications used for security testing to breach Fortune 500 companies. An investigation revealed that over 1,900 vulnerable applications were exposed, allowing attackers to deploy crypto miners and webshells. Many of these apps used default credentials and lacked proper security measures.
Anthropic's report reveals that AI agents exploited vulnerabilities in smart contracts, simulating over $550 million in potential losses. They discovered new zero-day vulnerabilities, highlighting the urgent need for improved security measures in blockchain technology.
This article outlines various security risks associated with AI agents and their infrastructure, including issues like chat history exfiltration and prompt injection. It emphasizes the need for a comprehensive security platform to monitor and govern AI operations effectively.
Google released an urgent update for Chrome to fix two vulnerabilities that can be exploited by simply visiting malicious web pages. Users should ensure they're on version 143.0.7499.146 or later to stay protected from potential attacks.
Socket has launched a Threat Intel page that tracks ongoing supply chain attack campaigns affecting open-source packages. The new feature helps teams quickly determine if they are impacted by these coordinated attacks and provides context for affected packages.
Daniel, a 16-year-old hacker, details how he and friends discovered critical vulnerabilities in Mintlify, an AI documentation platform. They found a cross-site scripting flaw that could have allowed attackers to compromise accounts across several major companies, including Discord. After reporting the issue, they received bounties for their findings.
This article outlines the updated OWASP Top Ten list for 2025, highlighting critical web application security risks. It introduces two new categories and shifts existing ones based on survey data and trends in software vulnerabilities. Each category is defined by specific weaknesses, helping organizations focus on key security issues.
Portugal's revised cybercrime law creates a legal safe harbor for security researchers acting in good faith. Researchers can now engage in certain hacking activities without fear of prosecution, provided they meet specific conditions, such as reporting vulnerabilities promptly and not seeking financial gain.
Researchers found insecure bootstrap scripts in legacy Python packages that could allow attackers to exploit a domain takeover. The scripts fetch an outdated installation package from a now-available domain, which poses a risk of executing malicious code. Some affected packages have removed the scripts, but others, like slapos.core, still include them.
The article details a reverse engineering project on the TP-Link Tapo C200 camera, revealing multiple security vulnerabilities. The author used AI tools to assist in the analysis, uncovering issues like a memory overflow and an unauthenticated WiFi hijacking exploit that could compromise user privacy.
Meta's Bug Bounty Program marked its 15th anniversary, awarding over $4 million in bounties this year alone, totaling more than $25 million since its start. The program is expanding with a new pilot for experienced researchers and highlighting significant findings, including vulnerabilities in WhatsApp and Oculus.
This article outlines how Dux AI helps organizations manage security vulnerabilities by identifying exploitable risks and applying quick mitigations. It emphasizes the importance of acting swiftly to protect against threats in a rapidly changing environment. Dux aims to streamline the remediation process, allowing teams to focus on critical issues.
Researchers found serious security flaws in the LINE messaging app, allowing for message replay attacks, impersonation, and sensitive data leaks. Despite LINE's claims of low risk, the app's integral role in daily life across East Asia raises significant privacy concerns.
Microsoft is disabling the NT LAN Manager (NTLM) protocol by default in Windows 11 and Windows Server to enhance security. Despite its long history, NTLM is outdated and vulnerable to attacks, prompting the shift towards more secure protocols like Kerberos. Many organizations still use NTLM, but the risks now outweigh the benefits.
This article discusses vulnerabilities in large language model (LLM) frameworks, highlighting specific case studies of security issues like remote code execution and SQL injection. It offers lessons learned for both users and developers, emphasizing the importance of validation and cautious implementation practices.
This article reviews the rise of agentic browsers, AI tools that autonomously navigate and perform tasks online. It highlights security vulnerabilities these browsers face and outlines the defensive measures implemented by developers. The piece also discusses the ongoing debate about the balance between autonomy and access to sensitive data.
The article discusses six newly discovered JavaScript zero-day vulnerabilities that could allow attackers to exploit package managers and execute malicious code. Experts warn that these flaws could enable large-scale supply chain attacks, especially if attackers gain access to package maintainers' credentials. The need for stronger security measures in software supply chains is emphasized.
A security researcher revealed how attackers can exploit Anthropic's Claude AI by using indirect prompt injections to extract user data. By tricking Claude into uploading files to the attacker's account, sensitive information, including chat conversations, can be exfiltrated. The researcher reported this issue, but Anthropic initially dismissed it as a model safety concern.
The Go programming language has released updates 1.25.6 and 1.24.12 to fix six critical vulnerabilities, including denial-of-service risks and potential arbitrary code execution. Developers are urged to upgrade immediately to avoid exploitation in unpatched environments.
Mandiant has released rainbow tables that significantly simplify the process of cracking NTLMv1 passwords, allowing attackers to recover authentication keys in under 12 hours using consumer-grade hardware. The release transforms this previously theoretical vulnerability into a practical threat, requiring organizations to take immediate action to mitigate risks.
This article provides a comprehensive analysis of the Shai Hulud 2.0 supply chain attack, detailing the compromised code libraries and the extent of the breaches. It also lists tools and methods for detecting and mitigating the impact of these attacks, emphasizing the importance of version pinning and runtime monitoring.
Microsoft Copilot allows non-technical users to create AI agents easily, but this can lead to serious security vulnerabilities. A recent report shows how these agents can be manipulated to leak sensitive data and cause data exposure. The simplicity of deployment makes it easy for users to overlook necessary security measures.
Anthropic tested ten AI models on 405 smart contract exploits and found that they could replicate over half of them, generating $4.6 million in simulated attacks. The study highlights the speed at which AI can identify vulnerabilities, raising concerns about security in decentralized finance.
This article explains the differences between prompt injection and SQL injection, emphasizing that prompt injection poses unique risks in generative AI systems. It highlights the challenges in mitigating these vulnerabilities due to the lack of distinction between data and instructions in large language models.
Two critical 0-day vulnerabilities in NetSupport Manager allow attackers to execute code remotely without authentication. This exploit can compromise industrial control systems, enabling lateral movement within networks. Organizations are urged to upgrade to the latest version to mitigate risks.
XBOW is a platform that automates penetration testing, offering faster and deeper vulnerability assessments than traditional methods. It validates findings through real exploitation, allowing security teams to focus on actual risks rather than theoretical ones. This helps address the growing challenge of security in the face of increasing cyber threats.
Apple has released updates for macOS and other platforms, addressing 19 security vulnerabilities in WebKit. These flaws could allow for various attacks, including data leaks and privilege escalation. The company reports no known active exploits of these vulnerabilities.
The article discusses a benchmark report that highlights how Anthropic's Claude models excel in security compared to other large language models (LLMs). While most models struggle with vulnerabilities like jailbreaks and harmful content generation, Claude consistently demonstrates superior performance, indicating a significant gap in safety standards across the industry.
This article discusses security vulnerabilities found in Command & Conquer: Generals, particularly in its online multiplayer functionality. The authors detail a memory corruption issue that allows for remote code execution, demonstrating the exploit with a custom worm. They also provide insights into the game's network architecture and packet structure.
Apple released security updates addressing 105 vulnerabilities in MacOS 26.1 and 56 in iOS 26.1 and iPadOS 26.1. The updates fix flaws across multiple devices but lack detailed severity ratings, frustrating some security experts. No active exploitation of these vulnerabilities has been reported.
SolarWinds released patches for three critical vulnerabilities in its Serv-U file transfer solution. One flaw allows attackers with admin privileges to execute arbitrary code, posing significant risks to affected systems. The vulnerabilities are listed in the CISA's Known Exploited Vulnerabilities catalog.
Researchers have identified multiple severe security flaws in the Coolify self-hosting platform, including command injection vulnerabilities that could allow attackers to execute arbitrary code and gain root access. Users are urged to update to fixed versions immediately, as around 52,890 instances are currently exposed.
This article examines how well AI models Claude Code and OpenAI Codex can identify Insecure Direct Object Reference (IDOR) vulnerabilities in real-world applications. It reveals that while these models excel in simpler cases, they struggle with more complex authorization logic, leading to a high rate of false positives.
Echo, a cybersecurity firm founded by Eilon Elhadad and Eylam Milner, raised $35 million in Series A funding to enhance its vulnerability management for container images. The company's solution aims to eliminate security issues from the base image layer, promising immediate reduction in vulnerability counts for clients.
This article details multiple security vulnerabilities discovered in Mintlify's documentation platform, including remote code execution and cross-site scripting flaws. The author and collaborators successfully exploited these issues, leading to significant risks for Mintlify's clients, including major companies like Discord and Vercel. They also describe the swift response from Mintlify in patching these vulnerabilities.
Moltbook, a social network for AI agents, suffered a major security breach due to a misconfigured Supabase database, exposing 1.5 million API keys and personal data of 17,000 human users. The incident highlights risks in quickly developed applications without adequate security measures.
SecObserve has moved its repositories to a new organization and changed the location of its Docker images. This open-source tool helps software teams manage vulnerabilities and licenses, integrating with CI/CD pipelines for easier scanning and reporting.
GoBruteforcer is a botnet attacking cryptocurrency databases to brute-force user passwords for various services. Its operators exploit weak credentials and misconfigured servers to expand their control, utilizing a mix of common usernames and a persistent malware infrastructure. Recent activities also show attempts to identify blockchain accounts with funds.
Three serious vulnerabilities in the runC container runtime could allow attackers to bypass isolation and gain root access to the host system. The flaws affect multiple versions of runC, with potential exploits requiring the ability to configure custom mounts. While no active exploitation has been reported, developers recommend using mitigations like user namespaces and rootless containers.
Google Big Sleep has published reports detailing several fixed vulnerabilities in the Linux kernel. The post provides a link to the full list of issues they discovered and addressed. This information is relevant for anyone concerned about Linux security.
Researchers found that hackers are using the React2Shell vulnerability to compromise NGINX web servers, redirecting traffic for malicious purposes. This can lead to malware infections and damage to an organization's reputation. CSOs are advised to secure server configurations and apply the latest security patches.
A recent investigation revealed over thirty vulnerabilities in major AI-integrated IDEs, exposing them to data theft and remote code execution. The flaws stem from how AI agents interact with existing IDE features, creating new attack vectors that attackers can exploit. Immediate mitigations are possible, but a fundamental redesign of IDEs is necessary for long-term security.
Oligo Security has revealed an ongoing global hacking campaign, ShadowRay 2.0, where attackers exploit a flaw in the Ray AI framework to create a self-propagating botnet. The attackers, known as IronErn440, leverage AI-generated payloads to enhance their methods while competing with other criminal groups for resources. Over 230,000 Ray servers are currently exposed to this threat.
This article explores how differing syntax interpretations across programming languages can lead to security vulnerabilities. It outlines techniques for creating payloads that exploit these ambiguities, allowing attackers to bypass filters and perform unauthorized actions. Practical examples and case studies illustrate the potential impact of syntax confusion.
The Codacy AI Risk Hub helps teams enforce secure coding practices for AI-generated code. It prevents vulnerabilities by tracking model usage, scanning for security risks, and managing hardcoded secrets across projects. This tool aims to maintain code quality while leveraging AI capabilities.
This article analyzes security flaws in various NFC technologies, including MIFARE Ultralight C and NTAG 223/224 DNA. It reveals how attackers can exploit these weaknesses to recover cryptographic keys through relay attacks and memory manipulation, highlighting the risks in real-world deployments.
This article details a security audit of Outline, an open-source wiki, comparing manual testing with AI security platforms. The audit identified seven unique vulnerabilities, some linked to external libraries and others specific to the Outline codebase. Key issues included server-side request forgery and insecure content handling.
The article discusses the security vulnerabilities associated with OpenClaw AI, particularly as companies increasingly integrate AI agents into their workflows. Experts warn about prompt injection risks and the potential for unauthorized access to sensitive data, emphasizing the need for companies to adopt strict security measures.
This article discusses an AI-driven security platform that enhances application security by analyzing code, infrastructure, and business logic. It reduces false positives and offers actionable fixes directly within developers' workflows. The platform claims to improve the identification of real vulnerabilities while streamlining the remediation process.
Fog Security revealed methods to evade AWS Trusted Advisor's S3 security checks, allowing public access to S3 buckets without triggering alerts. Despite reporting these issues to AWS, initial fixes were incomplete, leading to continued inaccurate assessments of bucket security. Their communication regarding the problem's severity was also criticized as insufficient.
SAP addressed 19 security vulnerabilities, including a critical flaw in SQL Anywhere Monitor that allowed remote code execution through hardcoded credentials. Experts recommend discontinuing the use of SQL Anywhere Monitor and deleting existing database instances as a temporary fix. Other vulnerabilities in SAP Solution Manager and SAP NetWeaver AS Java were also patched.
This article discusses the creation of the Domain Name System (DNS) by Paul Mockapetris, highlighting its role in converting domain names to IP addresses. It points out the centralization of DNS as a vulnerability in an otherwise decentralized internet, referencing Tim Berners-Lee's concerns about this weakness.
Aikido Security has identified a vulnerability in GitHub Actions and GitLab CI/CD workflows that allows AI agents to execute malicious instructions, potentially leaking sensitive information. The flaw affects multiple companies and demonstrates how AI prompt injection can compromise software supply chains.
Microsoft will now reward researchers for identifying critical vulnerabilities in any of its online services, regardless of the code's origin. This change aims to enhance security by incentivizing the discovery of flaws in both Microsoft's own and third-party components that impact its services.
Attackers exploited vulnerabilities in SolarWinds Web Help Desk to steal high-privilege credentials from various organizations. Microsoft is investigating which specific flaws were used, as multiple recent and old CVEs are in play. Security teams are advised to apply patches and monitor for unauthorized remote management tools.
QNAP addressed seven zero-day vulnerabilities exploited at the Pwn2Own Ireland 2025 competition. The flaws affect multiple software products and operating systems, prompting QNAP to urge users to update their software and change passwords for better security. Additionally, a new version of QuMagie was released to patch a critical SQL injection vulnerability.
This article details security vulnerabilities in Airoha-based Bluetooth headphones that allow attackers to connect without authentication. It discusses three specific CVEs and their implications, including the potential for eavesdropping through compromised devices. Technical details and a verification tool are also provided for further research.
This article discusses a method for identifying software vulnerabilities by integrating large language models (LLMs) with static analysis tools like CodeQL. The authors highlight their tool, Vulnhalla, which filters out false positives and focuses on genuine security issues, illustrating the challenges of using LLMs in vulnerability research.
This article outlines key security vulnerabilities in Next.js applications, including SSRF, XSS, and CSRF. It provides practical tips and techniques for penetration testers to effectively assess Next.js apps.
Two vulnerabilities, named LookOut, discovered in Google Looker can lead to remote code execution and data exfiltration. Attackers with developer permissions can exploit these flaws to fully compromise Looker instances.
The article details a hands-on investigation of PostHog's security flaws, specifically focusing on server-side request forgery (SSRF) vulnerabilities. It outlines how these weaknesses allow for unauthorized access to internal services, culminating in an RCE chain through ClickHouse and SQL injection.
Amazon's Threat Intelligence team has halted operations by Russian GRU hackers targeting cloud infrastructure, particularly in the energy sector. The attackers shifted from exploiting software vulnerabilities to leveraging misconfigured network devices for access, prompting Amazon to take protective measures for affected customers and recommend security audits.
Trend Micro patched three vulnerabilities in its Apex Central management console, including a severe remote code execution flaw that could allow attackers to execute code as SYSTEM. The vulnerabilities were disclosed by Tenable and affect versions below Build 7190. Users are urged to apply the critical patch immediately.
Shannon is an AI tool designed to autonomously conduct penetration tests on web applications. It identifies vulnerabilities by executing real exploits, not just alerts, helping teams secure their code continuously rather than waiting for annual tests. This approach closes the security gap that arises from frequent code deployment.
This article presents a security reference designed to help developers identify and mitigate vulnerabilities in AI-generated code. It highlights common security anti-patterns, offers detailed examples, and suggests strategies for safer coding practices. The guide is based on extensive research from over 150 sources.
The article discusses a major security incident where 30,000 public Postman workspaces exposed sensitive information like API keys and tokens. It highlights the risks faced by various industries, real-world consequences of these leaks, and the factors leading to such vulnerabilities.
Day Two of Pwn2Own Automotive 2026 featured intense competition among security researchers, resulting in 29 new exploits and $439,250 awarded. The event has now totaled 66 unique vulnerabilities and over $955,000 in prize money. Fuzzware.io leads the standings as the contest nears its final day.
This article discusses two critical vulnerabilities found in Chainlit, an open-source framework for chatbots. These flaws could allow attackers to access sensitive files and take over cloud accounts, highlighting the distinct security risks of interconnected AI systems.
The article reports on 884 new Known Exploited Vulnerabilities (KEVs) identified in 2025, highlighting that nearly 29% were exploited on or before their CVE publication date. It emphasizes the rapid pace of exploitation and the need for organizations to prioritize timely remediation of both new and existing vulnerabilities.
Cydome has identified a new variant of the Mirai botnet, called Broadside, which exploits a vulnerability in TBK DVR devices used in maritime logistics. This variant not only conducts DDoS attacks but also attempts to harvest system credentials, posing a significant threat to shipping operations.
This article details how the Escape research team identified over 2,000 vulnerabilities in more than 5,600 applications built with vibe coding platforms. It explains their methodology, which included data gathering, attack surface scanning, and the introduction of the Visage Surface Scanner to analyze frontend code for security weaknesses.
This article details how an indirect prompt injection in Google's Antigravity code editor can exploit vulnerabilities to steal sensitive data from users. It describes the process by which malicious code can bypass security settings and exfiltrate credentials through a browser subagent. The piece highlights Google's acknowledgment of these risks and the inherent dangers of using the software without proper safeguards.
This article argues that AI integration in cybersecurity can create more vulnerabilities rather than enhance security. It highlights how hype around AI often overshadows the real risks, such as data leaks and poorly integrated systems, which can lead to significant security breaches.
Multiple critical flaws in the n8n open-source workflow platform allow authenticated users to execute arbitrary code on the server. Despite a fix being released, researchers found a bypass that could lead to complete control over n8n instances, exposing sensitive data and connected accounts. Users are urged to update to the latest version and review their workflows for security.
Chainguard's report highlights the significant security risks associated with less popular container images, which account for most vulnerabilities. While popular images like Python and Node are commonly used, the majority of critical issues exist in the long tail of images, emphasizing the need for better management and remediation practices.
Researchers assessed AI models' abilities to exploit smart contracts, revealing significant potential financial harm. They developed a benchmark, SCONE-bench, that demonstrates AI's capacity to discover vulnerabilities and generate exploits, emphasizing the need for proactive defenses.
Freedom Chat recently addressed two serious security issues that exposed users' phone numbers and PIN codes. A researcher discovered that phone numbers could be guessed and that PIN codes were visible to others in public channels, prompting a reset of all user PINs and an update to the app.
AI models like Claude Sonnet 4.5 can now execute complex multi-stage attacks on networks using standard open-source tools, eliminating the need for custom toolkits. This advancement allows AIs to exploit known vulnerabilities quickly, emphasizing the urgent need for timely security updates.
This article analyzes the security of over 20,000 web applications generated by large language models (LLMs). It identifies common vulnerabilities, such as hardcoded secrets and predictable credentials, while highlighting improvements in security compared to earlier AI-generated code.
New security flaws in React Server Components can lead to denial-of-service attacks and expose source code. Users must quickly patch their systems, as many remain vulnerable despite previous updates. Ongoing exploitation attempts are reported, particularly from attackers in North Korea and China.
Tenzai has introduced an AI-driven platform that conducts penetration testing to identify and fix vulnerabilities in enterprise software. Backed by $75 million in funding, the service aims to automate and scale the work of elite hackers, addressing the talent shortage in cybersecurity.
The article details eight vulnerabilities in Claude Code that allow arbitrary command execution without user approval. It explains how flaws in the permission model and regex blocklists can be exploited through various commands like `man`, `sort`, and `git`. Each method demonstrates a different oversight in command argument filtering.