Blahalla

Hackers bypass SonicWall VPN MFA due to incomplete patching

www.bleepingcomputer.com · 2 months ago

Threat actors brute forced VPN credentials and bypassed multi factor authentication on SonicWall Gen6 SSL VPN appliances, gaining internal access and deploying ransomware related tools within roughly 30 to 60 minutes. The vulnerability is tracked as CVE-2024-12802 and is caused by missing MFA enforcement for the user principal name login format, meaning Gen6 devices remain vulnerable unless administrators perform a specific manual LDAP reconfiguration after installing the firmware update, while Gen7 and Gen8 devices are fully remediated by firmware alone. Investigations between February and March show attackers performing reconnaissance, credential reuse, and attempts to deploy a Cobalt Strike beacon and a vulnerable driver, with logs that made the failed MFA bypasses appear as normal, and Gen6 appliances reached end of life on April 16, 2026 so migration to supported devices is recommended.
Investigating unauthorized access to GitHub-owned repositories

github.blog · 2 months ago

On Monday, May 18, a compromise of an employee device involving a poisoned Visual Studio Code extension published by a third party was detected and contained. The malicious extension version was removed, the endpoint was isolated, and incident response began immediately. The current assessment is that the activity involved exfiltration of internal repositories only, with the attacker’s claim of approximately 3,800 repositories directionally consistent with the investigation and no evidence of impact to customer information stored outside those internal repositories. Critical secrets were rotated starting Monday into Tuesday with the highest-impact credentials prioritized, logs are being analyzed and monitoring continues, and a fuller report will be published and affected customers notified if additional impact is discovered.
Microsoft Disrupts Malware-Signing Service Used by Ransomware Gangs

www.techrepublic.com · 2 months ago

Fox Tempest operated a malware signing as a service that abused Microsoft Artifact Signing to issue short lived Microsoft issued code signing certificates, producing more than one thousand certificates and operating hundreds of Azure tenants to make malicious software appear legitimate. The service ran through a portal called signspace dot cloud and was marketed via Telegram and other channels for thousands of dollars per transaction, enabling distribution of trojanized installers and signed malware families including Oyster, Vidar, Lumma Stealer and Rhysida ransomware. Microsoft’s Digital Crimes Unit disrupted the infrastructure in May 2026, revoked the attributed certificates, and released detections and mitigation guidance for defenders to block Fox Tempest enabled attacks.
Cyber Pros Can't Decide If AI Is a Good or a Bad Thing

www.darkreading.com · 2 months ago

There is widespread ambivalence among cybersecurity professionals: in a global ISC2 survey of 16,029 respondents, AI was identified as the emerging technology most likely to have both the greatest positive impact on organizations ability to secure themselves and the greatest negative impact on their security outlook. Respondents cited AI powered social engineering and high fidelity deepfakes as the top threats, with 52 percent rating AI among the most negative developments and agentic AI and quantum computing following at 34 percent and 32 percent respectively. At the same time many organizations are adopting AI, with 41 percent testing or evaluating it, 28 percent already AI integrated, 63 percent reporting significant productivity gains, and two thirds expecting AI to create more technical and communications jobs in cybersecurity. Industry observers warn that enthusiasm for AI has prompted risky or rushed integrations and superficial rebranding that are creating new security problems.
GitHub Confirms Breach, 4K Internal Repos Stolen

www.darkreading.com · 2 months ago

There was a breach at GitHub in which a threat actor known as TeamPCP exfiltrated approximately 3,800 internal repositories and advertised about 4,000 private repos for sale. The compromise began with a poisoned Visual Studio Code extension on an employee device, and GitHub removed the malicious extension, isolated the endpoint, initiated incident response, and rotated critical secrets while validating rotations and monitoring for follow on activity. TeamPCP is financially motivated, has repeatedly targeted the open source ecosystem including publishing the Shai Hulud worm source, and GitHub will publish a fuller report once its investigation is complete.
Fake Android Apps Commit Carrier Billing Fraud for Premium Services

www.darkreading.com · 2 months ago

There are nearly 250 fake Android apps targeting users in Malaysia, Thailand, Romania, and Croatia that covertly enroll victims in premium carrier billed services. They rely on WebView automation, JavaScript injection, abuse of Google SMS Retriever API to capture one time passwords, interception of OTPs, automated SMS sending, and browser session hijacking, and they sometimes disable Wi Fi to force cellular billing. The campaign began in March 2025 and remained highly active through at least the second week of January 2026, using three malware variants that automate subscription workflows, steal session cookies, and report activity in real time to operators via Telegram. Targeted carriers include DiGi, Celcom, Maxis, and U Mobile in Malaysia, and the operators distribute fake apps through TikTok, Facebook, and Google while showing harmless web pages on non targeted carriers to avoid detection.
OpenAI Prepares to File for an I.P.O. in Coming Weeks

www.nytimes.com · 2 months ago

OpenAI is preparing for an initial public offering that could value the company at up to one trillion dollars and may file with securities regulators as soon as the second half of 2026 though some planning points to 2027. The offering is intended to raise capital for massive AI infrastructure and expansion as the company pursues rapid growth, with an annualized revenue run rate reported near twenty billion dollars even as losses and missed targets have raised investor scrutiny. The company completed a restructuring to create a more traditional corporate structure that would allow a public listing and plans to reserve a portion of IPO shares for individual investors.
Mini Shai Hulud: Compromised @antv npm packages enable CI/CD credential theft

www.microsoft.com · 2 months ago

There is an active supply chain attack that compromised an antv maintainer account to publish malicious npm package versions that propagated through dependency chains into widely used libraries including echarts for react. The malicious payload is a heavily obfuscated JavaScript file that executes during npm install via a preinstall hook on Linux GitHub Actions runners and is designed to steal credentials from GitHub, Amazon Web Services, HashiCorp Vault, npm, Kubernetes, and 1Password while scraping runner process memory and performing privilege escalation. The malware exfiltrates data over encrypted HTTPS command and control channels and via the Git data API, forges SLSA provenance attestations to appear legitimate, and has propagated into thousands of projects increasing the risk of token and cloud credential theft. GitHub removed hundreds of malicious packages and invalidated tens of thousands of npm tokens, and recommended mitigations include reviewing dependency trees, pinning known good versions, running npm install with ignore scripts enabled, rotating credentials, and auditing CI CD logs and repositories.
Processes & Culture Top Reasons Behind Data Breaches

www.darkreading.com · 2 months ago

An analysis of 2024 breaches affecting Massachusetts residents found that processes and culture, including underreporting and limited visibility, remain primary drivers behind incidents. Identity and access management gaps such as weak passwords and missing multi factor authentication, together with insufficient patch management and internet facing vulnerabilities, were cited as common intrusion vectors. State officials emphasized that timely transparency and stronger controls are needed, noting measures like mandated multi factor authentication, 90 day password rotations, and compulsory annual training while warning that threat actors continue to exploit social engineering such as text message impersonations.
Microsoft Open-Sources RAMPART and Clarity to Secure AI Agents During Development

thehackernews.com · 2 months ago

There are two new open source tools from Microsoft called RAMPART and Clarity to help developers test the safety and security of AI agents. RAMPART is a Pytest native testing framework that lets engineers write and run safety and security tests against agents, covering adversarial and benign issues such as cross prompt injections, behavioral regressions, and data exfiltration, and evaluates outcomes via adapters while building on the earlier PyRIT tool. Clarity is a structured thinking partner that guides teams through problem clarification, solution exploration, failure analysis, and decision tracking so issues can be addressed early and red team learnings can be turned into reproducible, verifiable engineering assets.
Patch Now: Critical Flaw in OT Robot OS Gives Attackers Control

www.darkreading.com · 2 months ago

A critical command injection vulnerability in the operating system for collaborative robots allows an unauthenticated attacker with network access to the Dashboard Server to execute commands and gain administrative control of robotic controllers, risking operational disruption and physical safety. The flaw is tracked as CVE-2026-8153, has a CVSS 3.1 base score of 9.8, and Universal Robots released a patch in PolyScope version 5.25.1 while recommending mitigations such as disabling the Dashboard Server if unused, restricting access to trusted hosts, and segmenting OT from IT. No known exploitation has been reported and disclosure was coordinated with CISA and CERT/CC.
Cisco Nexus 3000 and 9000 Series Switches Border Gateway Protocol Denial of Service Vulnerability

sec.cloudapps.cisco.com · 2 months ago

There is a vulnerability in the Border Gateway Protocol enforce first as feature of Cisco Nexus 3000 Series Switches and Cisco Nexus 9000 Series Switches running in standalone NX OS mode that could allow an unauthenticated remote attacker to trigger BGP peer flaps and cause a denial of service. The vulnerability is caused by incorrect parsing of a transitive BGP attribute and can be exploited by sending a crafted BGP update through an established BGP peer session, which may cause the device to drop the session and flap with the forwarding peer. Cisco has released software updates and documented workarounds to address the issue and the vulnerability is tracked as CVE-2026-20171.
Cisco Secure Workload Unauthorized API Access Vulnerability

sec.cloudapps.cisco.com · 2 months ago

There is a critical vulnerability in the access validation of internal REST APIs in Cisco Secure Workload that could allow an unauthenticated remote attacker to access site resources with Site Admin privileges. The issue is caused by insufficient validation and authentication of REST API endpoints, and a crafted API request could let an attacker read sensitive information and make configuration changes across tenant boundaries. Cisco has released software updates to address the issue, there are no available workarounds, and the vulnerability is tracked as CVE-2026-20223.
Cisco ThousandEyes Enterprise Agent BrowserBot Command Injection Vulnerability

sec.cloudapps.cisco.com · 2 months ago

A vulnerability in the BrowserBot component of Cisco ThousandEyes Enterprise Agent could allow an authenticated remote attacker to execute arbitrary commands on Agents within the BrowserBot container as the node user due to insufficient input validation of user supplied command arguments. Exploitation requires valid ThousandEyes SaaS credentials and the ability to manage transaction tests. Cisco has addressed the issue in the ThousandEyes service, no customer action is required, there are no workarounds, the Security Impact Rating is Medium, and the issue is tracked as CVE-2026-20206.
Cisco ThousandEyes Virtual Appliance Authenticated Remote Code Execution Vulnerability

sec.cloudapps.cisco.com · 2 months ago

A vulnerability in the SSL certificate handling of Cisco ThousandEyes Virtual Appliance could allow an authenticated remote attacker to execute commands on the underlying operating system as the root user. The flaw is caused by insufficient validation of user supplied input and can be exploited by uploading a crafted certificate, but exploitation requires valid administrative credentials. Cisco has released software updates to address the vulnerability. There are no available workarounds, the security impact is rated Medium, and the issue is tracked as CVE-2026-20199.
Securing the gaming culture of cultures

www.microsoft.com · 2 months ago

There are more than 500 million monthly active players on Xbox consoles, PC, handheld, and cloud gaming, and more than three billion people engage with gaming globally, requiring protection of players, developer studios, platforms, intellectual property, live operations, and community trust. Different parts of the gaming ecosystem present distinct risks such as account takeover and phishing on platforms, fraud and in-game economy abuse, high-value unreleased intellectual property and insider or social engineering threats against studios, and configuration drift and overlapping permissions in central studio teams. Security is framed as collaboration over control, using identity and access management, compliance and governance tooling, threat intelligence and detection, and secure development lifecycles to enable low-latency, frictionless player experiences while maintaining resilience.
GitHub admits major source code leak after 3,800 internal repositories breached

www.csoonline.com · 2 months ago

There was a major GitHub security breach detected on May 19 in which attackers exfiltrated code from about 3,800 internal repositories after compromising an employee device with a poisoned Visual Studio Code extension; GitHub removed the malicious extension, isolated the endpoint, and said it is analyzing logs, validating secret rotation, and monitoring for follow-on activity. The TeamPCP group claimed responsibility, posted a list of breached repositories on LimeWire, and threatened to sell or publicly leak the stolen code unless a buyer paid at least fifty thousand dollars. Security firms link the incident to a broader supply chain campaign that backdoored the Nx Console VS Code extension, exposed thousands of developers for a brief window and advised updating from version 18.95.0 to 18.100.0, and that same campaign also pushed hundreds of malicious npm package versions across projects such as AntV and TanStack Router while harvesting credentials for Kubernetes, npm, AWS, 1Password, private keys, and GitHub.
Grafana breach caused by missed token rotation after TanStack attack

www.bleepingcomputer.com · 2 months ago

A single GitHub workflow token that was missed during post-incident rotation after the TanStack npm supply-chain compromise allowed attackers to access Grafana’s private repositories, with the intrusion tied to the Shai-Hulud credential-stealing campaign attributed to TeamPCP. Grafana detected malicious activity on May 1, 2026, rotated a large number of workflow tokens but missed one token that the attackers used to exfiltrate source code and operational business contact information. Current evidence indicates no customer production systems or production data were compromised, the codebase was not modified, and users are not required to take action at this time.
Quantum Bridge Raises $8 Million for Quantum-Safe Key Distribution Solution

www.securityweek.com · 2 months ago

Quantum Bridge, a Toronto company founded in 2019, raised eight million dollars in a Series A round, bringing its total funding to sixteen million dollars with participation from Wayra Telefónica, Cadenza VC, Club degli Investitori angel investors, HPE and Bacchus Venture Capital. There is a quantum safe key distribution platform built around a patented Distributed Symmetric Key Establishment protocol that uses pre shared random data and a secret sharing scheme so no single Security Hub holds the final key, providing protection against classical and quantum attacks. The Symmetric Key Distribution System packages that protocol with post quantum cryptography and quantum key distribution into a crypto agile solution that integrates into existing networks via an Ansible based installer and can be deployed as network appliances or endpoint based systems for government, defense, finance, critical infrastructure and communications.
Microsoft Rolls Out Mitigations for ‘YellowKey’ BitLocker Bypass

www.securityweek.com · 2 months ago

Microsoft rolled out mitigations after a zero day tracked as CVE-2026-45585 with a CVSS score of 6.8 was shown to allow a BitLocker bypass when an attacker with physical access boots into recovery from a USB containing the publicly released YellowKey exploit. The exploit places an FsTx directory on the USB and uses Transactional NTFS replay to delete winpeshl.ini in the Windows Recovery Environment, causing WinRE to spawn a command shell that exposes the underlying partition while BitLocker appears unlocked. Mitigations prevent the FsTx Auto Recovery utility autofstx.exe from automatically running when WinRE launches, and defenders are instructed to mount the WinRE image, remove autofstx.exe from the image registry hive, remount the image, and reestablish BitLocker trust, with Microsoft also recommending adding a BitLocker PIN. Researchers warned that the ability for a System Volume Information FsTx directory on one volume to modify another volume during replay may itself be a separate vulnerability of concern and one researcher claims YellowKey can bypass TPM plus PIN protections.
Introducing RAMPART and Clarity: Open source tools to bring safety into Agent development workflow

www.microsoft.com · 2 months ago

There are two open source tools designed to bring safety into agent development: RAMPART, a pytest-based continuous testing framework that encodes adversarial and benign scenarios as repeatable CI tests to capture issues like cross-prompt injection, and Clarity, a structured design-time tool that guides teams through problem clarification, failure analysis, and decision tracking. RAMPART supports statistical trials for probabilistic LLM behavior, integrates red team findings as runnable regression tests, and evaluates tool invocations and side effects, while Clarity records conversations as human-readable artifacts in a clarity-protocol directory, tracks staleness, and can generate review packets. Both projects are available now as open source and aim to make AI safety a continuous engineering discipline by turning red-team lessons and incident findings into reproducible tests and living artifacts.
CISA Contractor Exposed Sensitive Credentials in Public GitHub Repository

www.techrepublic.com · 2 months ago

CISA is investigating after a contractor’s public GitHub repository exposed AWS GovCloud administrative credentials, plaintext passwords, cloud keys, authentication tokens, logs, and internal build and deployment files. The repository named Private-CISA was created on November 13, 2025, included a file called importantAWStokens that contained credentials to three GovCloud servers, and showed commit logs indicating GitHub’s secrets detection feature had been manually disabled while many passwords followed easy to guess patterns. CISA took the repository offline after being notified, says there is no indication that sensitive data was compromised as the agency investigates, and a US senator has requested an urgent classified briefing on the incident.
Score by collisions, patch by panic: defensive architecture for the post-90-day-disclosure era

blog.himanshuanand.com · 2 months ago

There is a proposal to score vulnerability severity by collision count, raising severity and shrinking disclosure windows when multiple independent reports, working exploits, or public proofs of concept exist. Researchers are urged to ship patches with reports and vendors are urged to redesign systems assuming exploits may land before patches, emphasizing strict egress controls, ephemeral infrastructure rebuilt every twelve to twenty four hours, rootless runtime sandboxes, and system call filtering. Operational best practices include pinning dependencies and stopping automatic updates, continuous runtime validation and prebuilt virtual patching and WAF rules, written zero day playbooks and feature flags to quickly isolate or disable risky integrations. Independent researchers should assume they are not the only finder and push for short windows while defenders must automate safe fast deploy rails such as canaries, auto rollback and blue green deployments to close the current automation gap that favors offense.
AI-Powered App Attacks Are Faster, More Frequent and Harder to Stop

www.securityweek.com · 2 months ago

Agentic AI is collapsing the distinction between emerging targets and primary targets, enabling attackers to strike mobile apps across every industry within hours of release. Digital.ai's 2026 App Security Threat Report shows attacks against client facing apps rose from fifty five percent in 2022 to eighty seven percent in 2026, and iOS attack rates have converged with Android so that iOS now faces ninety seven percent of the prior Android rate. AI assisted reverse engineering, exploit generation, and dynamic analysis have made complex verticals easier to attack, driving steep increases in attacks against automotive apps and an eight percentage point rise against medical device apps. Defenders are urged to treat every app as a primary target from publication and adopt defensive agentic AI rather than relying on geographic distance or delayed detection.
[ Removed by Reddit ]

www.reddit.com · 2 months ago

NULL
CVE-2026-45585: Windows BitLocker — YellowKey Recovery Bypass Analysis

www.zerodaybrief.blog · 2 months ago

CVE-2026-45585, nicknamed YellowKey, allows an attacker with physical access to bypass BitLocker by placing crafted Transactional NTFS FsTx recovery log files on a USB stick that are replayed by the Windows Recovery Environment. When the system is booted into WinRE and the logs are replayed, WinRE can delete the winpeshl.ini recovery UI and instead present a SYSTEM command shell with the BitLocker volume mounted, while the encryption and TPM remain intact and no passwords are required. Microsoft released mitigations on May 20, 2026, recommending switching from TPM-only protectors to TPM plus PIN, and the issue affects Windows 11 24H2, 25H2, 26H1 and Windows Server 2025.
On AI Security

www.schneier.com · 2 months ago

Benchmarks cannot reliably measure AI capabilities or security because these are systemic properties that metrics alone fail to capture. Over the past thirty years, software security evolved from black box penetration testing through whitebox code analysis and architectural risk analysis to process driven assurance frameworks such as the Building Security In Maturity Model. A similar process oriented measurement approach will likely be needed for AI, while immediate progress can come from cleaning up outstanding work and applying robust assurance processes to manage risk. There is no single security meter for AI, so organizations must remain extra vigilant and adopt rigorous risk management and assurance practices.
Identity Alone Isn't Enough: Why Device Security Has to Share the Load

www.bleepingcomputer.com · 2 months ago

Identity checks alone are no longer sufficient because attackers increasingly use sophisticated phishing and real time proxying to capture session tokens and bypass multi factor authentication, allowing legitimate credentials to be replayed from attacker controlled devices. Zero Trust guidance such as NIST Special Publication 800 207 recommends factoring device posture into access decisions, and a defensible model combines continuous device verification, binding access to approved hardware, proportionate enforcement, and user self service remediation. Continuous device verification that monitors encryption, endpoint protection, patch status, and configuration across the session reduces the value of stolen credentials and intercepted tokens and can be operationalized by device trust solutions across Windows, macOS, Linux, and mobile platforms.
CVE-2026-34472: Pre-auth credential exposure and auth bypass in ZTE H188A V6 routers

minanagehsalalma.github.io · 2 months ago

A routing flaw in the ZTE ZXHN H188A V6 web interface allows unauthenticated requests to reach pre login wizard handlers and return WLAN, PPPoE, and administrative credentials, and this issue has been assigned CVE 2026 34472. The root cause is that empty root path requests can override routing by supplying query parameters named underscore type and underscore tag so the QuickSetupEnable gate is skipped, enabling a single unauthenticated GET to retrieve sensitive fields including the Wi Fi passphrase which on this device becomes the default administrator password when uppercased. The vulnerability enables full admin takeover when the management UI is reachable, was observed against roughly 500 publicly exposed interfaces, and as of May 18, 2026 the vendor characterized the issue as low risk and no public patch was available.
Grafana GitHub Security Incident Reportedly Connected to TanStack npm Ransomware

gbhackers.com · 2 months ago

Grafana Labs disclosed a targeted GitHub security incident linked to the TanStack npm supply chain ransomware campaign, detected on May 11, 2026, with attackers issuing a ransom demand on May 16, 2026. Attackers exploited a compromised GitHub Actions workflow token to access and download portions of public and private source code, internal operational repositories, and business contact information, with no evidence of code tampering or impact to customer systems or the Grafana Cloud platform. The incident persisted because at least one workflow token was overlooked during initial token rotation, allowing attackers to maintain access and continue exfiltration. Grafana rotated workflow and automation tokens, audited commits and repository activity since May 11, 2026, enhanced monitoring and telemetry, hardened continuous integration and continuous delivery pipelines, and notified federal law enforcement.
Iran Wants to Tax the Internet Flowing Through the Strait of Hormuz While Restricting Its Own Citizens Online

irannewswire.org · 2 months ago

Iran is considering imposing licensing fees and transit charges on submarine fiber optic cables passing through the Strait of Hormuz, targeting major cloud and technology companies including Meta, Alphabet and Microsoft. The proposal is presented as a new revenue stream amid severe economic pressure and a collapsing rial and would affect dozens of cables, including SeaMeWe-5 and AAE-1, that form a global digital chokepoint. The initiative contrasts with domestic policy that enforces systematic filtering, targeted throttling, network shutdowns during protests, and reliance on expensive or state linked VPN services that restrict citizens access to the global internet. Maritime lawyers and shipping industry analysts warn the plan would face legal resistance under international maritime law, but potential inspections or interference by naval vessels could still raise insurance premiums and alter risk calculations for international telecommunications consortia.
1Password Teams With OpenAI to Stop AI Coding Agents From Leaking Credentials

www.securityweek.com · 2 months ago

1Password and OpenAI have introduced an Environments MCP Server for OpenAI Codex that provides AI coding agents with just in time, scoped credentials while keeping secrets out of prompts, source code, repositories, terminals, and model context. Credentials remain end to end encrypted in 1Password vaults, are mounted into the application process memory only for the authorized process, require user authentication at access, and are discarded after use to maintain central governance and auditability. The approach is intended to prevent agent custody of persistent secrets, reduce exposure to prompt injection and exfiltration, and serve as an access architecture for future agentic systems.
How CISOs Should Prep for Agentic-Ready AI BOMs

www.darkreading.com · 2 months ago

There is growing recognition that AI bills of materials must document not only components such as models, datasets, frameworks, and dependencies but also runtime attributes that describe autonomous action. Agentic ready AI BOMs need both artifact lineage and authority lineage, including who an agent operates as, tool permissions, network egress constraints, action level authorization, and auditing. Research and prototype schema extensions to CycloneDX and SPDX show that adding runtime execution evidence improves reproducibility and vulnerability assessment accuracy. Security leaders are advised to treat AI systems as products, keep registers of models, data links, tool calls and owners, define approved behavioral baselines and drift, and constrain action spaces so every action is permissioned, attributable, and auditable.
Critical ExifTool Vulnerability Lets Hackers Compromise Macs via Malicious Images

gbhackers.com · 2 months ago

A newly disclosed vulnerability in ExifTool, tracked as CVE-2026-3102, allows attackers to execute arbitrary shell commands on macOS by embedding malicious payloads in image metadata fields such as FileCreateDate and MDItemFSCreationDate. The flaw affects ExifTool version 13.49 and earlier, is triggered when metadata is processed with the n flag and copied via the tagsFromFile feature so that an unsanitized date value reaches a system call, and echoes a prior input sanitization issue tracked as CVE-2021-22204. ExifTool 13.50 mitigates the issue by switching to argument based execution and adding a secure wrapper, and users are urged to update to version 13.50 or later, verify for embedded or outdated copies in third party software, and process untrusted files in isolated environments.
Anthropic Silently Patches Claude Code Sandbox Bypass

www.securityweek.com · 2 months ago

A vulnerability in Claude Code’s network sandbox allowed an attacker to bypass the outbound allowlist using a SOCKS5 hostname null byte injection and could be chained with prompt injection to exfiltrate environment variables, credentials, tokens and other infrastructure data. A related sandbox bypass was assigned CVE-2025-66479 and arose when the sandbox misinterpreted a block all outbound setting as allow everything, and that issue was fixed on November 26, 2025. The null byte bypass was present from Claude Code general availability on October 20, 2025 until fixes were committed on March 27, 2026 and shipped in Claude Code 2.1.88 on March 31, 2026, and the researcher criticized Anthropic for not assigning a CVE to this flaw or mentioning it in release notes.
Drupal critical update to fix bug with high exploitation risk

www.bleepingcomputer.com · 2 months ago

A core security release is scheduled for May 20 between 17:00 and 21:00 UTC, and Drupal warns that threat actors could develop exploits within hours of the disclosure, so administrators running Drupal 8 or 9 are strongly recommended to upgrade to at least 10.6 and reserve time to apply the update. The vulnerability affects Drupal core versions 8 and later though not all configurations are impacted, and security updates will be provided for Drupal 11.3.x, 11.2.x, 11.1.x, 10.6.x, 10.5.x and 10.4.x with guidance to update to Drupal 11.1.9 and 10.4.9 where applicable. Drupal will not publish technical details before the announcement, end of life versions 8 and 9 will receive no formal patches but hotfix files will be released for 9.5 and 8.9 to mitigate issues for sites on 9.5.11 or 8.9.20, and administrators are urged to monitor the official security portal and avoid unverified information.
Webworm Deploys EchoCreep and GraphWorm Backdoors Using Discord and MS Graph API

thehackernews.com · 2 months ago

There is fresh activity from a China aligned threat actor known as Webworm in 2025 that deployed two custom backdoors named EchoCreep and GraphWorm which use Discord and the Microsoft Graph API respectively for command and control. Webworm has been active since at least 2022 and has targeted government agencies and enterprises in IT services, aerospace, and electric power across countries including Russia, Georgia, Mongolia, Belgium, Italy, Serbia, Poland, and South Africa. EchoCreep supports file upload, file download, and remote command execution via cmd.exe, while GraphWorm can spawn new cmd.exe sessions, execute processes, upload and download files to and from Microsoft OneDrive, and stop its own execution on operator signal. The actor has shifted toward proxy and semi legitimate utilities, using a GitHub repository impersonating a WordPress fork, SoftEther VPN for blending in, and open source tools such as dirsearch and nuclei to brute force and scan web servers, with other observed tools including WormFrp, ChainWorm, SmuxProxy, and WormSocket.
Gremlin Stealer Hides C2 and Exfiltration Paths in Encrypted Resources

gbhackers.com · 2 months ago

A newly identified variant of the Gremlin stealer conceals its command and control and data exfiltration logic inside encrypted dot NET resource sections, using single byte XOR encoding and staged loading so malicious components are decrypted and executed only in memory to evade static analysis. It targets browser stored credentials, session tokens, cryptocurrency wallets, clipboard contents, VPN and FTP credentials, and Discord tokens, compresses stolen artifacts into ZIP archives labeled with the victim public IP, and exfiltrates them to attacker controlled infrastructure. The campaign adds clipboard hijacking that replaces wallet addresses, WebSocket based session hijacking, expanded Chromium and in memory data targeting, and uses commercial packing with instruction virtualization along with identifier renaming, string encryption, and control flow obfuscation to frustrate detection and analysis.
Microsoft DurableTask Python Client Targeted in TeamPCP Cyberattack

gbhackers.com · 2 months ago

The TeamPCP supply chain campaign compromised the official Microsoft DurableTask Python client by using a compromised GitHub account to extract a PyPI token and publish trojanized durabletask releases 1.4.1, 1.4.2, and 1.4.3 to PyPI. The malicious payload, delivered as rope.pyz and injected into package modules, collects credentials from AWS, Azure, GCP, Kubernetes, Vault, local files and password managers, attempts brute force against password managers and propagates via AWS Systems Manager and Kubernetes to additional hosts. Responders are urged to scan dependency files and CI logs for the affected versions, search Linux systems for artifacts such as /tmp/managed.pyz and the infection marker under ~/.cache/.sys-update-check, rotate all cloud and application credentials, audit CloudTrail and Kubernetes logs, and block identified command and control domains.
NVIDIA Triton Inference Server Flaw Raises Risk of Unauthorized Access

gbhackers.com · 2 months ago

There is a critical authentication bypass vulnerability in NVIDIA Triton Inference Server tracked as CVE-2026-24207 with a CVSS v3.1 score of 9.8 that can allow remote unauthenticated attackers to gain unauthorized access. Successful exploitation can lead to unauthorized code execution, privilege escalation, tampering of machine learning models and outputs, denial of service, and exposure of sensitive data in GPU accelerated inference environments. Versions prior to r26.03 are affected and users are advised to upgrade to r26.03 or later, implement network segmentation and strong authentication, enable logging and anomaly detection, and monitor systems for suspicious activity.
Agent AI is Coming. Are You Ready?

thehackernews.com · 2 months ago

On May 19, 2026, Orchid Security released Identity Gap Snapshot 2026 and found identity dark matter now represents 57 percent of identity assets compared with 43 percent visible. Agent AI are described as shortcut seekers that, when blocked, can use hard coded credentials stored in plaintext, borrow higher privilege credentials, or grab broadly accepted tokens to complete tasks. Top findings include that two of every three nonhuman accounts are created locally and remain unmanaged, 70 percent of applications have excessive privileged accounts, and 40 percent of accounts are orphaned. Well managed identity and access management is presented as a critical control, with organizations urged to use readiness checklists and act now to contain autonomous AI activity.
SHub Reaper impersonates Apple, Google, and Microsoft in one MacOS attack chain

www.csoonline.com · 2 months ago

A newly disclosed macOS infostealer campaign called SHub Reaper impersonates Apple, Google, and Microsoft at different stages of a single attack chain to trick users into executing malicious AppleScript through Script Editor and evade Terminal paste protections. The attack uses fake Apple security alerts to lure victims into launching Script Editor with pre populated malicious scripts, then performs environment and anti analysis checks, deploys additional payloads, and establishes persistence via LaunchAgents that masquerade as legitimate vendor files. Once active, it harvests browser credentials, password managers, Keychain data, cryptocurrency wallets including MetaMask and Phantom, messaging applications, and user documents. Defensive guidance recommends monitoring unexpected Script Editor and osascript activity and spawned processes, investigating suspicious LaunchAgents and typo squatted domains, and implementing strict web filtering.
GitHub Breached — Employee Device Hack Led to Exfiltration of 3,800+ Internal Repos

thehackernews.com · 2 months ago

GitHub is investigating unauthorized access to roughly 4,000 internal repositories after threat actor TeamPCP listed the platform’s source code and internal organization data for sale for no less than 50,000 dollars. GitHub says there is currently no evidence of impact to customer information stored outside its internal repositories and will notify customers if any impact is discovered. TeamPCP’s Mini Shai Hulud campaign also compromised the durabletask PyPI package by publishing malicious versions 1.4.1, 1.4.2, and 1.4.3 after obtaining a PyPI token from a compromised GitHub account; the package contains a dropper that fetches a second stage payload and installs a Linux-only infostealer able to harvest cloud credentials, password manager vaults, SSH keys, Docker credentials, VPN configurations, and other secrets. The malware propagates using stolen tokens via AWS SSM and kubectl exec, uses a mechanism that retrieves backup command and control data from public commit messages, and any machine or pipeline that installed an affected package version should be treated as fully compromised.
GitHub hit by a compromised VSCode extension

xcancel.com · 2 months ago

GitHub's internal repositories were breached by a malicious Visual Studio Code extension. Microsoft had previously closed a request to add update cooldowns as not planned but may reconsider the control following the breach. A new request has been opened in the Visual Studio Code project to pursue update cooldowns again.
GraphWorm Malware Abuses Microsoft OneDrive for Stealthy C2 Operations

gbhackers.com · 2 months ago

There is new Webworm activity during 2025 showing a shift from Asian targets to government entities across Europe and activity in South Africa, and the group has introduced a OneDrive based backdoor called GraphWorm that uses the Microsoft Graph API for command and control. GraphWorm creates a unique victim identifier from system attributes, places each victim in a dedicated OneDrive directory with job and result subfolders, encrypts communications with AES 256 CBC and Base64, uses the createUploadSession endpoint for large file exfiltration, and supports proxy configuration to evade monitoring. The actor also deployed EchoCreep using Discord for C2, leveraged custom proxy tools including WormFrp ChainWorm SmuxProxy and WormSocket, abused misconfigured cloud storage and GitHub repositories for staging, and exploited known flaws including CVE-2017-7692 in SquirrelMail for initial access in at least one observed case.
Over 320 NPM Packages Hit by Fresh Mini Shai-Hulud Supply Chain Attack

www.securityweek.com · 2 months ago

There are over 320 NPM packages affected by a fresh Mini Shai Hulud supply chain attack after a maintainer account named atool with access to the antv namespace was compromised and used to publish malicious package versions that propagated to popular downstream packages and CI environments. The install time payload reads GitHub Actions runner memory to extract masked CI CD secrets, harvests credentials from more than 130 file paths covering cloud providers, Kubernetes, HashiCorp Vault, cryptocurrency wallets and developer tools, and exfiltrates stolen data via GitHub repositories and a fallback server while also dropping persistent backdoors and downloading Python code to enable remote execution. Researchers tracked over one thousand malicious package versions across ecosystems with NPM representing the overwhelming majority, and the malware can abuse registry APIs to validate tokens, enumerate and modify packages, add preinstall hooks, bump versions and republish under compromised identities, with additional compromises observed on PyPI and in a GitHub Action and attribution linked to TeamPCP.
Fox Tempest Linked to Malware-Signing Service Abusing Microsoft Artifact Signing

gbhackers.com · 2 months ago

Fox Tempest operated a large scale malware signing as a service that abused Microsoft's Artifact Signing platform to generate fraudulent but valid short lived code signing certificates that made trojanized installers appear trusted. The service created more than one thousand certificates and hundreds of Azure tenants, moved from a web portal to preconfigured virtual machines for customers, and monetized tiered pricing while enabling threat clusters including Vanilla Tempest and Storm variants to distribute payloads such as Rhysida ransomware, the Oyster backdoor, Lumma Stealer, and Vidar infostealer. Microsoft revoked the certificates and disrupted the operation with its Digital Crimes Unit and partners in May 2026, and recommended layered defenses including cloud delivered protection, Safe Links and Safe Attachments, tamper protection, attack surface reduction rules, SmartScreen browser protections, and use of Security Copilot for hunting and response.
Caught Off Guard: Securing AI After It Hits Production

www.securityweek.com · 2 months ago

Enterprises rushing AI projects into production are forcing security teams into reactive mode because security is often left out of early development and successful experiments are moved to production without oversight. To improve readiness, organizations should use data driven discussions to engage application owners, increase security agility to operate across complex hybrid and multi cloud environments, and strengthen operational workflows to integrate AI data and alerts quickly. They should future proof existing application and API security stacks so AI layer protections can be added without rebuilding, practice proactive hygiene including continuous scanning for vulnerabilities and sensitive data exposures, and deploy contextual awareness capabilities to detect runtime attacks, abuse, fraud, and DDoS against the AI layer in near real time. Taken together, these strategic steps can help security teams respond more quickly and appropriately when AI applications are thrust into production with little warning.
Fake Tax Assessment Pages Spread Windows Malware

gbhackers.com · 2 months ago

A campaign tracked as TAXTRIDENT is using fake Indian income tax assessment pages to trick Windows users into downloading ZIP archives or script downloaders that deliver a signed ClientSetup installer or a signed ManageEngine UEMS agent. One delivery path drops a ZIP with a signed ClientSetup executable that creates a hidden client directory, installs services and drivers such as a MANC service and fake svchost processes, writes YTSysConfig configuration files, and opens persistent command and control channels, while alternate paths use an AssessmentOrder.vbs downloader or PHP-named endpoints that return VBScript and stage further payloads. Defenders are advised to hunt for behavioral indicators including script execution from unusual sources, misleading file extensions, unexpected service or driver creation, UAC policy changes, and outbound connections to unknown servers rather than relying solely on file hashes.
Exploit released for new PinTheft Arch Linux root escalation flaw

www.bleepingcomputer.com · 2 months ago

A publicly available proof of concept exploit targets a recently patched Linux local privilege escalation vulnerability named PinTheft and can grant local attackers a root shell on Arch Linux. PinTheft abuses an RDS zerocopy double free that can be converted into a page cache overwrite via iouring fixed buffers and requires the RDS kernel module to be loaded, iouring enabled, a readable SUID root binary, and x8664 support for the payload. Administrators are urged to install the latest kernel updates or mitigate risk by unloading or blacklisting the RDS module, and PinTheft is still awaiting assignment of a CVE identifier for tracking.