GitHub is investigating unauthorised access to internal repositories after TeamPCP used a malicious Visual Studio Code extension on an employee device, exposing thousands of repositories. Separate npm supply-chain compromises affected the @antv ecosystem and art-template package, introducing malicious scripts and iOS browser exploit delivery infrastructure.

Linux kernel CVE-2026-46333 allows local privilege escalation through ptrace flaws, enabling credential disclosure and root execution across major distributions since 2016. ExifTool CVE-2026-3102 enables macOS command execution via crafted image metadata, affecting versions 13.49 and earlier through system call abuse in user context.

Webworm expanded European targeting, deploying EchoCreep and GraphWorm backdoors using Discord and Microsoft Graph API for command-and-control operations in 2025. Microsoft released mitigations for CVE-2026-45585 YellowKey BitLocker bypass, modifying WinRE and BootExecute to block exploitation paths requiring physical access attack scenarios.

Verizon DBIR 2026 reports vulnerability exploitation has overtaken credential theft as primary breach vector, driven by ransomware and third-party compromises. 7-Eleven confirmed a breach affecting franchisee document systems after claims by ShinyHunters of unauthorised access to internal environments during investigation phase. Android Premium Deception campaign used hundreds of fake apps to silently enrol users in paid services, generating unauthorised recurring charges.

GitHub on Tuesday said it is investigating unauthorised access after TeamPCP listed internal repositories for sale following employee device compromise via malicious VS Code extension. GitHub reported no evidence of customer data impact outside internal repositories, while assessing claims of 3,800–4,000 repos and rotating critical secrets. Separate campaign linked to TeamPCP involved compromise of Microsoft durabletask PyPI package delivering Mini Shai-Hulud infostealer targeting cloud credentials, SSH keys, and developer tools.

Microsoft updated CVE-2026-45585 guidance detailing mitigation steps for BitLocker bypass "YellowKey", involving WinRE modification and BootExecute autofstx.exe removal from Session Manager registry hive, after mounting WinRE image. Procedure includes mounting WinRE via reagentc, loading SYSTEM hive, editing HKLM\WinREHive\ControlSet001\Control\Session Manager BootExecute to remove autofstx.exe, then unloading hive and committing changes, re-enabling WinRE trust. Additional mitigations advise enforcing TPM-and-PIN BitLocker protection using manage-bde, PowerShell, Intune, or Group Policy, restricting TPM-only configurations vulnerable to physical access exploitation.

Kaspersky disclosed CVE-2026-3102 in ExifTool affecting macOS versions 13.49 and earlier, where crafted image metadata can trigger arbitrary command execution via system calls. Exploitation leverages unsanitised FileCreateDate handling and DateTimeOriginal copying via -tagsFromFile, while -n enables raw values that reach system() execution through metadata parsing pathway. CVE allows attacker-controlled images to execute shell commands with user privileges, potentially enabling full system compromise; vulnerability was discovered by Kaspersky GReAT.

Palo Alto Networks Unit 42 tracked TamperedChef-style campaigns distributing trojanised productivity applications via malvertising, mapping clusters CL-CRI-1089, CL-UNK-1090 and CL-UNK-1110 with overlapping code signing reuse. Palo Alto Networks identified over 4,000 samples and around 100 variants, linked to 12,000 observed instances, using 80+ code-signing entities and advertising transparency data. Activity clusters deliver infostealers, RATs and adware components with delayed activation, persistence via scheduled tasks or registry run keys secondary payload retrieval after dormancy periods.

ESET Research analysed 2025 activity of China-aligned APT Webworm, tracking shift from Asian targets towards European organisations and evolving intrusion toolsets. In campaigns, Webworm deployed EchoCreep and GraphWorm backdoors using Discord and Microsoft Graph API for command and control, alongside GitHub-staged malware delivery and proxy tools. Operators decrypted over 400 Discord messages and used compromised infrastructure including Amazon S3 buckets and custom multi-hop proxies to manage victim systems.

CVE-2026-46333 Linux kernel __ptrace_may_access flaw allows unprivileged local users to disclose sensitive files and execute arbitrary commands as root across major distributions. Researchers reported the issue has existed since Linux v4.10-rc1 in 2016 and can be chained with pidfd_getfd to escalate privileges via four tested targets. Exploits demonstrated credential disclosure and root execution against chage, ssh-keysign, pkexec and accounts-daemon across Debian, Ubuntu, Fedora systems with widespread kernel exposure since 2016.

The art-template npm package was compromised after a repository takeover introduced malicious versions injecting external JavaScript loaders into bundled template files. The injected code loads remote domains that redirect to a watering-hole framework targeting Safari on iOS 11.0 through 17.2, resembling Coruna exploit delivery. The payload implements anti-bot checks, WebAssembly-based fingerprinting, and continuous beaconing to a command and control endpoint while executing staged remote module loading.