Attacks & Vulnerabilities |
Over 14,000 F5 BIG-IP APM instances still exposed to RCE attacks (2 minute read)
CVE-2025-53521, initially disclosed in October 2025 as a DoS bug in F5 BIG-IP APM, was later reclassified as a critical remote code execution (RCE) vulnerability actively being exploited after new details emerged in March. Attackers without authentication are targeting systems with access policies on virtual servers. Shadowserver reports over 17,100 internet-visible BIG-IP APM systems, with more than 14,000 still unpatched, despite CISA including the flaw in its Known Exploited Vulnerabilities (KEV) catalog and instructing federal agencies to fix the issue by Monday. F5 recommends that defenders review disk, logs, and terminal history for indicators of compromise, consider UCS backups from potentially affected systems as untrusted, and rebuild compromised systems from trusted, clean sources. |
Critical Vulnerability in Claude Code Emerges Days After Source Leak (4 minute read)
Anthropic accidentally published a debug sourcemap for Claude Code v2.1.88 to npm, exposing 512,000 lines of TypeScript code across 1,900 files, which is now permanently replicated online. Separately, Adversa AI found a flaw in Claude Code's permission system: feeding it a 50+ subcommand pipeline via a malicious CLAUDE.md file silently disables all deny rules, with no user warning. This opens the door to stealing SSH keys, AWS credentials, and GitHub tokens, as well as poisoning CI/CD pipelines. |
Money transfer app Duc exposed thousands of driver's licenses and passports to the open web (3 minute read)
Toronto-based Duales left an Amazon S3 bucket for the money transfer app Duc publicly accessible, no password required, exposing over 360,000 files, including driver's licenses, passports, selfies, names, addresses, and transaction records dating back to September 2020. Duales CEO Henry Martinez González confirmed it's fixed, but won't say whether access logs exist, and now Canada's privacy regulator is investigating. |
|
STARDUST CHOLLIMA Likely Compromises Axios npm Package (4 minute read)
On March 31, North Korean threat actor STARDUST CHOLLIMA used stolen maintainer credentials to compromise the Axios npm package, introducing a trojanized version containing cross-platform ZshBucket variants targeting various operating systems. This version featured a JSON-based command-and-control (C2) protocol and enabled payload injection and remote execution. Defenders should audit npm credentials and treat Axios installations from this date onward as potentially compromised, given the actor's focus on currency generation and fintech supply chains. |
Mutation testing for the agentic era (6 minute read)
MuTON and mewt are open source mutation testing tools built for agentic workflows. MuTON targets TON blockchain languages (FunC, Tolk, and Tact), while mewt provides language-agnostic support for Solidity, Rust, Go, and more. Both use Tree-sitter for AST-aware multi-line mutation and SQLite for persistent, resumable campaign state. The tooling improves on prior approaches like slither-mutate and universalmutator by enabling mutant prioritization, SARIF output, and flexible filtering, making AI-assisted triage token-efficient. Security teams auditing smart contracts should evaluate MuTON and mewt alongside a companion configuration-optimization skill to right-size campaign runtimes and surface blind spots that code-coverage metrics routinely miss. |
Google Workspace's continuous approach to mitigating indirect prompt injections (4 minute read)
Google's GenAI Security Team has developed a layered defense approach to combat indirect prompt injection (IPI) in Workspace using Gemini, which includes human red-teaming, automated attack generation, and a centralized vulnerability registry. They expand discovered attack methods into various forms through the Simula synthetic data pipeline to enhance machine learning model training and improve security measures. AI application developers should adopt a similar multi-layered strategy, combining fixed controls with ongoing model retraining and comprehensive assessments to prevent security regressions. |
|
Introducing EmDash — the spiritual successor to WordPress that solves plugin security (11 minute read)
Cloudflare's EmDash (v0.1.0) is an open source, MIT-licensed CMS built in TypeScript on Astro and Cloudflare Workers designed to replace WordPress' fundamentally insecure plugin architecture. Each plugin runs in an isolated Dynamic Worker sandbox with capabilities declared statically in a manifest, limiting blast radius to only explicitly granted permissions rather than full database and filesystem access as in WordPress. EmDash deploys to any Node.js server or Cloudflare, includes built-in passkey authentication, a remote MCP server, x402 payment support, and WordPress import tooling. |
Linx Security (Product Launch)
Linx Security offers an identity security platform that maps and monitors human, non-human, and agentic identities across enterprise environments, using real-time detection and automated remediation to cut manual oversight and close lifecycle blind spots. |
Zerobox (GitHub Repo)
Lightweight, cross-platform process sandboxing powered by OpenAI Codex's runtime. Sandbox any command with file, network, and credential controls. This is a new tool that was created within the last two weeks and should be used with caution. |
|
Linx Security Raises $50 Million for Identity Security and Governance (1 minute read)
Linx Security raised a $50 million Series B led by Insight Partners alongside Cyberstarts and Index Ventures, bringing total funding to $83 million for its AI-native identity security and governance platform. Founded in 2023, the New York-based startup maps, monitors, and governs human, non-human, and agentic identities across enterprise environments through its Autopilot AI agent, which provides real-time threat detection and automated remediation with minimal manual oversight. The funding will be directed toward product development, go-to-market expansion, and growing its global footprint. |
Fake Claude Code source downloads actually delivered malware (2 minute read)
According to Zscaler's ThreatLabz, a GitHub repository from user idbzoomh surfaced as a top Google search result for "leaked Claude Code," masquerading as a leaked TypeScript source for Anthropic's CLI. However, the available .7z downloads actually contained a Rust-based dropper that deployed Vidar v18.7 to harvest browser history, credit card information, and credentials. The infection also included GhostSocks, a tool that repurposes compromised systems into proxy nodes. |
No Paste for You! Reverse Engineering Apple's ClickFix Protections (5 minute read)
Patrick Wardle reverse-engineered the ClickFix protection in macOS 26.4, linking it to two undocumented Endpoint Security events and revealing details like source and target processes along with clipboard contents. This protection requires a System Integrity Protection check, which prevents third-party tools from accessing the same kernel-level authentication used by Apple. Without a public Endpoint Security paste event, defenders must rely on less reliable keystroke monitoring methods. |
|
Microsoft Warns of WhatsApp Attachments Spreading Backdoor on Windows PCs (2 minute read)
Microsoft Defender researchers warned of an active campaign since late February delivering VBS files via WhatsApp that chain through renamed LOLBins (curl.exe as netapi.dll, bitsadmin.exe as sc.exe), pull second-stage payloads from AWS S3, Tencent Cloud, and Backblaze B2, disable UAC via registry modification, and install unsigned remote access tools such as AnyDesk.msi to establish persistent backdoor access. |
|
| Love TLDR? Tell your friends and get rewards! | | Share your referral link below with friends to get free TLDR swag! | | | |
| Track your referrals here. |
|
| |
0 Comments