Unit 42 reports a structural shift in the cyber extortion economy: encryption-based ransomware dropped to 78% of cases in 2025 (from 90%+ in prior years), with pure data-theft extortion rising shar...

Cisco Talos disclosed four heap-based buffer overflow vulnerabilities (CVE-2026-25104, CVE-2026-25713, CVE-2026-28764, CVE-2026-22554) in MediaArea’s MediaInfoLib version 26.01, a widely-used open-...

A Black Hills ANTISOC red team engagement narrative demonstrating how a helpdesk technician’s habit of reusing a single predictable password for all resets created a systemic vulnerability. Operato...

Red Canary investigates threat detection for Microsoft Entra Agent ID — a new identity class for autonomous AI agents distinct from human and workload identities. The article focuses on a scenario ...

EvidenceForge is a new open-source tool from Cisco Talos that generates correlated synthetic security logs across 20+ formats (Windows Security Events, Sysmon, EDR/XDR, Zeek, Snort, Linux syslog, e...

CVE-2026-23918 is a high-severity double-free vulnerability in Apache 2.4.66’s mod_http2 module, triggered by a HEADERS frame followed immediately by an RST_STREAM frame with a non-zero error code ...

Doyensec conducted a structured side-by-side evaluation of two AI-powered penetration testing platforms — Aikido’s Attack AI Pentest and XBOW’s Lightspeed — against real web applications. The asses...

This article explores applying module stomping (DLL hollowing) to Position-Independent Code (PIC) and PIC Objects (PICOs) within the Crystal Palace framework. The classic technique loads a legitima...

Red Canary’s May 2026 threat intelligence roundup covers the top 10 most prevalent threats observed in April. ClearFake leads via compromised-website JavaScript injection using fake CAPTCHA/ClickFi...

Merchants accepting card payments over VoIP cannot use the reduced-scope SAQ C-VT or P2PE Self-Assessment Questionnaires to limit their PCI DSS compliance burden — a fact many have only discovered ...

Aura InfoSec presents a Claude Code CLI plugin that performs semi-autonomous AWS cloud security reviews by combining AI-driven reasoning with active environment probing. Unlike rule-based scanners ...

Sansec discovered CVE-2026-45247 (CVSS 9.8), an unauthenticated PHP object injection vulnerability in Mirasvit Cache Warmer for Magento/Adobe Commerce. The extension deserializes a client-controlle...

A pentest engagement against a multi-tenant social media monitoring/PR platform uncovered a cross-tenant data exposure via an atypical path traversal. The asset library download endpoint accepted f...

The TeamPCP npm supply chain campaign (wave 5 in the post-Shai-Hulud series) hit Alibaba’s @antv data visualization suite via 323 malicious packages with ~16M weekly weekly downloads; payloads grep...

This Doyensec post analyzes AWS Application Load Balancer (ALB) misconfigurations that create unintended access paths. Three concrete bug classes are detailed: (1) CloudFront/WAF bypass when the or...

This article covers defensive remediation of Kerberoasting in Active Directory environments. Kerberoasting abuses standard Kerberos behavior: any authenticated domain user can request a service tic...

Vulnerability Spoiler Alert is an LLM-based tool that monitors open-source git repositories for security patches before CVEs are published. The key insight is framing the problem as ‘does this comm...

HTB MonitorsFour write-up chaining multiple vulnerabilities on a Windows host. PHP type juggling (loose comparison flaw) bypasses authentication on an API endpoint, leaking crackable password hashe...

A practitioner-oriented guide from Fortbridge on securing enterprise LLM deployments, covering four primary control areas: RAG pipeline security and access controls, input/output validation and fil...

This weekly roundup combines several security items, but the most technically useful sections cover a new SHub “Reaper” macOS stealer variant and two actively exploited Microsoft Defender zero-days...

Fox-IT analyzes a Lazarus-linked three-stage Windows implant chain made up of DPAPILoader, RemotePELoader, and RemotePE, where the final RAT is executed entirely in memory and never written to disk...

Unit 42 tracks an Iran-linked Screening Serpens espionage campaign that used tailored job-recruitment phishing lures against aerospace, defense, telecom, and technology targets in the U.S., Israel,...

Trail of Bits describes hardening zizmor, a static analyzer for GitHub Actions workflows, after GitHub’s addition of YAML anchors exposed parsing and analysis gaps in a tool meant to catch CI/CD mi...

Unit 42 analyzes how attackers operationalize ROADtools against Microsoft Entra ID, focusing on two core modules: ROADrecon for tenant enumeration and roadtx for token acquisition, exchange, and de...

This article analyzes how guardrail placement inside an AI agent’s orchestration loop changes the ability to stop indirect prompt injection attacks, using a demo where a benign request to summarize...

The post introduces Striga, a Python-based x86_64 lifter that translates machine instructions into LLVM IR to make binary analysis and transformation easier than working directly on assembly. Its c...

TrustedSec documents a renewed Shai-Hulud npm supply-chain campaign in which a compromised maintainer account published malicious versions of more than 300 packages across the @antv ecosystem, plus...

Datadog analyzes CVE-2021-25740, an unpatchable Kubernetes design issue where a tenant with permission to modify Endpoint or EndpointSlice objects can repoint a Service to arbitrary IPs instead of ...

Unit 42 maps multiple TamperedChef-style malware clusters by correlating reused code, shared code-signing infrastructure, corporate registration data, and advertising overlap across more than 4,000...

Splunk’s May 2026 UBA advisory is a dependency remediation bulletin rather than a root-cause write-up, documenting security fixes delivered by upgrading or removing dozens of bundled third-party co...

Splunk’s advisory for AppDynamics Private Synthetic Agent 26.4.0 is a bundled dependency remediation notice rather than a root-cause write-up: the release updates or removes a large set of embedded...

Datadog describes how it publishes OpenVEX documents for its own agents, container images, and packages so downstream users can suppress SCA findings that are present in dependency metadata but not...

The article frames modern influence operations as a cybernetic control problem and argues that large-scale collection of human telemetry can be paired with algorithmic feedback loops to shape perce...

Wiz documents a TeamPCP supply-chain compromise involving the PyPI package durabletask, where a trojanized package release turned a normal Python dependency install into an attacker foothold. The k...

Cisco Talos disclosed a batch of patched vulnerabilities spanning TP-Link Archer AX53 firmware, Adobe Photoshop’s Microsoft Store installer, OpenVPN, and Norton VPN, with the most serious issues en...

Cisco Talos analyzes a BadIIS variant identified by embedded “demo.pdb” paths and argues it operates as a commodity malware ecosystem shared or sold among multiple Chinese-speaking crime groups rat...

TrustedSec proposes Coverage-Driven Sustained Testing (CDST), a graph- or matrix-based orchestration model for agentic security testing that generates new work from uncovered areas instead of stopp...

Doyensec shows that NASA CFITSIO’s Extended Filename Syntax (EFS) is effectively a filename mini-language processed by ffopen(), where strings are reinterpreted into protocol handlers, output claus...

Quarkslab describes a pre-authentication takeover chain against VSOL GPON OLTs and the vendor’s Cloud EMS fleet manager that could let an attacker move from an exposed access device to control of a...

ETH Zurich’s Phoenix research demonstrates the first successful Rowhammer-induced bit flips on DDR5 DRAM protected by modern Target Row Refresh (TRR) mitigations, showing that DDR5 is not immune to...

SentinelOne analyzes a new SHub Stealer variant, “Reaper,” that delivers a macOS infostealer through fake WeChat and Miro installers while rotating its disguise across Apple, Microsoft, and Google ...

This issue is a curated AWS-focused security roundup rather than original research, but it highlights several technically relevant items for cloud defenders. The strongest entries include an Amazon...

Datadog Security Labs introduced Pathfinding Labs, a collection of more than 100 intentionally vulnerable AWS environments designed to reproduce IAM privilege-escalation paths, CSPM misconfiguratio...

This Hack The Box write-up chains multiple real-world style bugs, starting with an unauthenticated directory traversal in Pterodactyl Panel’s locale endpoint that lets PHP include attacker-controll...

SentinelOne surveys how attackers are increasingly compromising the software delivery pipeline itself rather than only poisoning downstream dependencies, focusing on build servers, self-hosted runn...

Unit 42 analyzes a newer Gremlin Stealer variant that hides its second-stage payload inside the .NET Resource section, XOR-encodes the embedded data, and in some samples adds commercial packer prot...

This write-up examines macOS design-level security flaws that can be chained into more serious outcomes than a standalone userland root bug, focusing on how TCC and SIP remain the real security bou...

Cisco Talos reports active in-the-wild exploitation of multiple Cisco Catalyst SD-WAN flaws, most notably CVE-2026-20182, an authentication bypass in SD-WAN Controller and Manager that lets an unau...

TrustedSec presents a technical deployment guide for passkeys that focuses less on basic WebAuthn concepts and more on the operational security issues organizations hit during rollout. The article ...

Sansec reports active exploitation of a critical Funnel Builder by FunnelKit flaw affecting WooCommerce stores, where all versions before 3.15.0.3 expose a public checkout endpoint that lets unauth...

Datadog analyzed three malicious node-ipc npm releases (9.1.6, 9.2.3, and 12.0.1) that embedded a backdoor in the package’s CommonJS entrypoint, so the payload executed on require("node-ipc") rathe...

Datadog traces a coordinated supply-chain campaign that poisoned 170 legitimate npm and PyPI packages and links it to a separate compromise of the official Cemu GitHub release assets. The malicious...

SentinelOne summarizes telemetry from more than 11,000 environments and argues that unmanaged AI credentials are becoming a distinct cloud-risk class, with exposed OpenAI and Azure OpenAI API keys ...

Praetorian shows how single-page applications leak useful attack surface data because unauthenticated visitors still receive the full JavaScript bundle, including route definitions, API endpoint UR...

The piece is a methodology article on finding new vulnerabilities rather than a disclosure of a specific bug. It recommends starting from an operational gap during red team work, reviewing prior ad...

This article shows how Linux cgroup metadata can be repurposed as a DFIR and detection signal to enrich process telemetry during server compromise investigations. It explains how cgroups v2 expose ...

Google Project Zero shows how its earlier Dolby zero-click bug (CVE-2025-54957) was adapted from Pixel 9 to Pixel 10 by retargeting library offsets and replacing the old __stack_chk_fail overwrite ...

Outpost24’s May 2026 Patch Tuesday roundup highlights 137 Microsoft fixes, including 30 Critical issues, and calls out several bugs with strong enterprise impact even though none are confirmed zero...

Datadog analyzes the leaked source code of Shai-Hulud, a modular TypeScript/Bun offensive framework attributed to TeamPCP that operationalizes software supply-chain poisoning, credential harvesting...

The post discloses a Teachable CDN misconfiguration where uploads.teachablecdn.com exposes the first page of an S3 bucket listing for lecture_attachments through CloudFront, even though direct acce...

This write-up covers GHSA-f9f8-rm49-7jv2, a Composer bug where GitHub OAuth token validation rejects GitHub’s newer ghs_<id>_<JWT> Actions token format because the regex only allows [.A...

Cisco Talos summarizes Microsoft’s May 2026 Patch Tuesday release, which fixes 137 vulnerabilities, including 31 rated critical, and calls out the issues most relevant for network defenders and det...

PortSwigger argues that end-to-end AppSec workflows such as vulnerability discovery, validation, and remediation are becoming automatable with frontier and open-weight LLMs, but that the limiting f...

The write-up describes an unauthenticated information disclosure bug in a SaaS platform where a simple GET request to /api/oauth/provider/ returned the full OAuth configuration for every configured...

ProjectDiscovery’s April 2026 Nuclei template release adds 226 templates covering 123 CVEs, including multiple CISA KEV entries and high-impact checks for auth bypass, RCE, SSRF, path traversal, SQ...

Trail of Bits introduces gosentry, a fuzzing-focused fork of the Go toolchain that preserves the standard testing.F and go test -fuzz workflow while replacing the native engine with a LibAFL-backed...

Cisco Talos outlines how state-sponsored intrusions differ from ransomware-style incidents by emphasizing covert, long-term access obtained through valid credentials, supply-chain compromise, or oc...

TrustedSec documents a recurring attack pattern where adversaries prime victims with phishing emails, then call them over Microsoft Teams while impersonating IT and walk them into starting a Window...

Wiz details a software supply-chain intrusion in which malicious versions of TanStack and other npm packages were published, turning routine dependency installation into attacker-controlled code ex...

Unit 42 analyzes how attackers abuse Active Directory Certificate Services by chaining certificate template misconfigurations, overly broad enrollment rights, and shadow credential techniques to im...

The alert ties observed intrusions involving the EtherRat remote-access trojan and TukTuk command-and-control infrastructure to a final-stage deployment of Gentleman ransomware. The technical value...

This digest curates several technically useful AWS security items rather than presenting a single original finding. The strongest material covers experiments showing LLM agents can turn a leaked AW...

Datadog Security Labs analyzes a supply-chain style attack against coding agents by abusing Claude Code skills, especially the ! dynamic-context feature that executes shell commands before the rend...

The post describes BaremetalVMM, a personal Firecracker-based microVM manager built largely with LLM assistance to support security testing workloads that do not fit well in containers. The tool ca...

This research dissects Visual Studio Code Dev Tunnels as a potential remote-access and C2 channel rather than just a developer convenience feature. The author shows that tunnel discovery starts wit...

This Hack The Box write-up chains several Windows attack techniques on a domain controller running SMB, MSSQL 2022, WinRM, and a .NET Message Framing/WCF service. The initial foothold comes from an...

The post analyzes newer ClickFix initial-access lures that replace the well-known Win+R execution path with Windows Power User Menu (Win+X) workflows and user-launched Terminal sessions to avoid Ru...

This weekly roundup aggregates three security developments: U.S. sentencing actions against a Karakurt extortion negotiator and two facilitators of DPRK laptop-farm operations, SentinelLABS reporti...

Wiz describes a Linux kernel local privilege-escalation bug class in the networking stack where packet fragment handling can be driven through the ESP (IPsec Encapsulating Security Payload) and RxR...

The post analyzes Audi’s myAudi connected-car platform by reversing the web and Android app flows, bypassing certificate pinning with Frida/Objection, and tracing the authentication chain across id...

The post is a technical postmortem on why AWS Bedrock Claude usage was showing as effectively $0 when filtered by a cost-allocation tag, despite more than $62 in real spend. The author traced the i...

The write-up describes an unauthenticated remote code execution issue in Fooocus, a widely used open-source AI image generation WebUI, caused by unsafe use of Python eval on user-controlled metadat...

This article breaks down Kubernetes secret handling as a set of concrete threat models: secrets at rest in etcd and on worker nodes, secrets in transit between components, and API-level attacks aga...

Certitude argues that current frontier models are already changing vulnerability research by automating most of the workflow, including idea generation, exploit prototyping, and report drafting; th...

The article outlines four common browser attack surfaces that accumulate through normal use: overprivileged extensions, long-lived sessions and stored credentials, phishing links, and drive-by down...

TrustedSec outlines a GRC-focused checklist for enterprise AI adoption, emphasizing that the main security failure modes are data leakage, insecure model output, third-party exposure, and unmanaged...

Unit 42 describes CVE-2026-0300, a buffer overflow in PAN-OS User-ID Authentication Portal (Captive Portal) that lets an unauthenticated attacker send crafted network traffic and achieve arbitrary ...

Project Black describes an authenticated exploit chain in the Gibbon school management platform v30.0.00 that starts with SQL injection in modules/Tracking/graphing.php and ends in code execution. ...

Quarkslab breaks down the Independent Guest Virtual Machine (IGVM) binary format used to package and securely launch a VM’s initial state, including firmware, kernel, and initrd, across different v...

This post releases a public corpus of 140+ benign weaponizable file samples intended to validate whether mail gateways, web proxies, and endpoint controls block risky formats in practice rather tha...

This post introduces Swapper, a lightweight Burp Suite extension that solves a common testing problem in stateful applications: requests that invalidate a session or anti-CSRF token after each use....

The post reports a critical buffer overflow in Palo Alto Networks PAN-OS that is already being exploited in the wild, making it immediately relevant for organizations exposing PAN-OS firewalls or V...

Cisco Talos analyzes telephone-oriented attack delivery (TOAD) campaigns by treating phone numbers in scam emails as durable indicators of compromise rather than focusing only on disposable sender ...

The article describes a Chrome bug that bypassed SameSite=Strict by abusing how DevTools re-fetched a cross-site POST response when a service worker was registered on the target origin. Opening Dev...

CVE-2026-31431 (“Copy Fail”) is a deterministic Linux local privilege escalation bug in the kernel’s AF_ALG AEAD path, specifically the algif_aead module combined with the authencesn algorithm and ...

The article analyzes a now-patched Anthropic behavior where a documented test token, ANTHROPIC_MAGIC_STRING_TRIGGER_REFUSAL_..., deterministically caused Claude 4 streaming responses to terminate w...

Trail of Bits walks through two C/C++ bug patterns from its testing handbook: a Linux ping utility that is reachable for command injection, and a Windows WDF driver handler that can be turned into ...

Cisco Talos profiles UAT-8302, a China-nexus espionage cluster that has targeted government entities in South America and southeastern Europe since at least late 2024, and ties it to a shared tooli...

Cisco Talos describes a 2026 intrusion chain where a fake ScreenConnect update launches a Rust dropper, which writes a .NET loader disguised as update.txt into ProgramData and persists it via a sch...

The post introduces pyghidra-mcp v0.2.0, which adds a --gui mode that lets the same MCP server controlling project-wide Ghidra analysis also drive a live CodeBrowser window. Instead of fighting Ghi...

TrustedSec argues that LLMs materially reduce the cost of reverse engineering defensive products, compressing analysis of five commercial endpoint tools from weeks to days by automating mapping, su...

Doyensec shows how multi-tenant AWS Cognito User Pools that let each tenant register its own external OIDC or SAML IdP can be abused when security checks are placed in the wrong Lambda triggers. Th...

Quarkslab and OSTIF audited Paramiko and its use of pyca/cryptography, focusing on SSHv2 key exchange, signature handling, entropy sources, constant-time behavior, and the security of the surroundi...

Antirez describes the four-month design and implementation of a new Redis Array type, focusing on a sparse array representation that can handle very large numeric indexes without huge allocations. ...

This AWS Security Digest issue is a curated roundup of several technically useful cloud-security findings rather than a pure product announcement. The most actionable item is AWS’s March 2026 Threa...

The article proposes a practical framework, DOVE (Detection Overlap & Value Evaluation), for reviewing whether custom threat-hunting and SIEM detections still add value or have become duplicate...

The post shows how ISC DHCP Server can be driven to unauthenticated root RCE by chaining intended features rather than exploiting memory corruption. If omapi-port is exposed without an omapi-key, a...

The article argues that detection content should be engineered with the same rigor as application code, proposing a Use Case Development Lifecycle (UCDL) for SIEM/XDR analytics instead of the commo...

The article argues that endpoint-only monitoring misses common multi-surface intrusion patterns and that detection workflows need correlated telemetry from identity, cloud, network, SaaS and unmana...

This weekly roundup is mostly news-oriented, but its strongest technical section covers the ‘Mini Shai-Hulud’ supply-chain attack against SAP-related npm packages such as @cap-js/attachments, @cap-...

The article proposes a detection-engineering model based on attacker/defender economics: some alerts are ‘rabbit-side’ events where a miss has high business impact, while others are ‘fox-side’ even...

PortSwigger outlines a methodology for turning manual pentest heuristics into Burp custom scan checks so the same detection logic can run across large application and API estates. The article focus...

This post analyzes enterprise audiovisual hardware used in meeting rooms and documents multiple serious flaws, starting with CVE-2026-26461 in the Aver PTC320UV2 camera. The author reversed the fir...

Unit 42 analyzed 18 malicious browser extensions posing as AI productivity tools and found they abused extension privileges to steal sensitive data, including email drafts, ChatGPT prompts, passwor...

TrustedSec shows that a principal with only WriteGPLink on an Active Directory OU can weaponize existing GPOs that reference UNC paths, without editing the GPO itself or modifying SYSVOL. The attac...

Quarkslab breaks down why Microsoft Entra ID application auditing is hard in practice by separating the roles of App Registrations and Service Principals, then tracing how effective access expands ...

Star Labs details CVE-2026-41873 in the retired Lua version of Apache Pony Mail, where CRLF injection in email.lua can be abused to smuggle arbitrary HTTP requests to the backend Elasticsearch serv...

Unit 42 describes AirSnitch, a novel enterprise Wi-Fi attack class that bypasses the confidentiality assumptions of WPA2/WPA3-Enterprise by exploiting protocol-infrastructure interactions rather th...

This article is an incident-response playbook for confirming and containing a suspected data breach rather than a vulnerability disclosure. It recommends validating the alert by checking source cre...

This post summarizes OX Security’s findings on the Anthropic Model Context Protocol (MCP) SDK, arguing that a protocol-level design choice can enable remote code execution across Python, TypeScript...

Cisco Talos’ Q1 2026 IR report highlights phishing as the top initial access vector again, including a case where attackers used Softr’s AI-driven app builder to generate credential-harvesting page...

Cisco Talos maps out macOS-native living-off-the-land techniques for execution, lateral movement, and payload staging, focusing on primitives that operate outside the SSH-centric telemetry many def...

Cisco Talos analyzes 2025 phishing and identity attacks, highlighting a shift from generic spam to workflow-themed lures such as IT, travel, expense, and logistics messages that are designed to cap...

Praetorian argues that vulnerability triage should prioritize exploit chains rather than standalone CVSS scores, showing how two individually scored browser bugs can combine into a realistic full-h...

This post is a technical walkthrough of how OAuth application consent actually works in Microsoft Entra ID, aimed at correcting common misconceptions that appear when teams deploy third-party Micro...

This post analyzes how Linux’s pipe_buffer object can be turned into multiple kernel exploit primitives by corrupting specific fields: flags for Dirty Pipe-style writes to read-only files, ops for ...

The post walks through a red-team operation that initially used HTML smuggling to deliver a 7z archive containing a signed executable and malicious DLLs, then relied on DLL hijacking to drop a payl...

Wiz frames CI/CD pipelines as a high-value attack surface because build jobs routinely run with repository write access, cloud credentials, registry tokens, SSH keys, and OIDC-issued identities, so...

This post describes a compromise of Context.ai OAuth tokens that let attackers operate through an already-trusted SaaS integration, turning delegated access into a supply-chain attack path. The cor...

Unit 42 argues that frontier AI models have crossed from being coding helpers into systems that can autonomously reason about vulnerability discovery, exploit chaining, and control bypass, which co...

The article explores Microsoft 365 threat hunting techniques that use the email InternetMessageId field as a detection signal instead of relying only on subjects, sender domains, or other higher-le...

The post proposes a measurement framework for detection engineering programs based on visibility and telemetry coverage rather than raw alert counts. It recommends tracking concrete baselines such ...

This AWS Security Digest issue is a curated roundup rather than original research, but it highlights two technically meaningful findings. The first is CVE-2026-5429 in AWS Kiro, where the IDE injec...

The post introduces swic, a minimal Go-based replacement for calibre-web that the author wrote after submitting 17 security-related pull requests to calibre-web and deciding its feature set created...

This write-up reverse engineers CVE-2025-20393 in Cisco Secure Email Gateway and Secure Email and Web Manager, showing how an unauthenticated attacker can reach remote code execution through the En...

Doyensec describes fuzzing the widely used CFITSIO FITS-parsing library and finding multiple memory-corruption issues, first with a generic AFL++ campaign and then by targeting the library’s Extend...

This Hack The Box write-up chains multiple wireless and web weaknesses in a simulated enterprise environment, starting with an SNMP-exposed default consultant password that grants SSH access to a c...

This post documents a small but concrete SQLite implementation detail: temporary files created during operations like VACUUM are prefixed with etilqs_, which is simply sqlite_ spelled backwards. VA...

This weekly roundup aggregates three technically relevant stories: the FBI and Indonesian authorities dismantled the W3LL phishing ecosystem, a phishing-as-a-service platform that cloned login port...

Trail of Bits shows that Google’s published zero-knowledge proof for optimized quantum cryptanalysis was forgeable because the Rust-based zkVM prover pipeline contained multiple memory-safety and l...

Sygnia frames Anthropic’s Mythos Preview and Project Glasswing as evidence that exploit development timelines may compress sharply, with claimed capabilities including zero-day discovery in major o...

TrustedSec argues that Mythos is a real capability jump for vulnerability research because it can identify promising bug candidates, reason about exploitability, and help produce exploit paths fast...

Unit 42’s updated threat brief tracks multiple Iran-linked campaigns tied to the 2026 regional conflict, including renewed OT/ICS targeting by CL-STA-1128 (Cyber Av3ngers/Storm-0784) against Rockwe...

This write-up shows how an attacker who already reached on-premises Active Directory Domain Admin can pivot into Microsoft 365 by stealing Entra ID session tokens from a privileged user’s workstati...

This post examines how real anonymous credential systems are built and deployed, focusing on Privacy Pass as a concrete implementation of Chaum-style blind-signature credentials. It walks through t...

Datadog analyzes common GitHub Actions misconfigurations that enabled recent CI/CD attacks, focusing on cases where user-controlled values like PR titles, commit messages, and branch names are inte...

Quarkslab examines how LLVM’s middle-end optimization passes can act as an automatic de-obfuscator by simplifying intentionally complex arithmetic and control-flow constructs back to their original...

Unit 42 analyzes active exploitation attempts against CVE-2023-33538, a command-injection flaw in the /userRpm/WlanNetworkRpm(.htm) endpoint on several end-of-life TP-Link routers where the ssid1 p...

Datadog argues for enforcing dependency “cooldowns” to reduce the risk of consuming freshly published malicious package versions, using the recent Axios npm compromise and prior s1ngularity/Shai-Hu...

Cisco Talos documents a previously unreported PowerShell botnet, “PowMix,” used in phishing campaigns against Czech organizations and job seekers, delivered through malicious ZIP archives containin...

Cisco Talos disclosed seven patched memory-safety issues across Foxit Reader and LibRaw, including Foxit Reader CVE-2026-3779, a use-after-free triggered by crafted JavaScript embedded in a malicio...

This post argues that AI-assisted vulnerability discovery does not scale like proof-of-work systems, because repeated sampling from a weaker model eventually saturates the useful execution paths an...

The article describes a proxy-execution technique against Microsoft Edge WebView2 Runtime in which trusted Windows Apps such as Teams, Outlook, and other Store-style apps spawn msedgewebview2.exe, ...

Red Canary’s recap distills several operational trends from its Threat Detection Report, emphasizing that attackers are increasingly targeting identity material such as credentials, OAuth consent g...

Praetorian outlines how “shadow admin” accounts in Active Directory gain Domain Admin-equivalent power through indirect control paths rather than membership in obvious privileged groups. The post h...

The article shows that on Windows domain controllers, setting LmCompatibilityLevel=5 on a non-PDC DC does not reliably block NTLMv1 authentication because the DC is not treated as authoritative for...

CVE-2026-20205 affects Splunk MCP Server app versions before 1.0.3 and stems from CWE-532-style logging of sensitive data, exposing user session tokens and authorization tokens in clear text. An at...

Splunk’s April 2026 advisory is a bundled dependency-remediation notice for Splunk Enterprise rather than a root-cause write-up, but it is still operationally useful because it maps affected produc...

This write-up covers a CTF challenge built around chaining an HTTP/2 request smuggling flaw with a Twig sandbox escape to achieve remote code execution. The first stage abuses H2 request desynchron...

This CTF write-up presents a multi-stage web exploit chain that starts with Nginx UNIX socket injection to reach an internal application surface that was not intended to be exposed directly. The ne...

This write-up analyzes CVE-2022-25967, an RCE in the Eta Node.js templating engine caused by unsafe propagation of Express render options into Eta’s template compilation config. Eta’s file handling...

This CTF write-up shows how a nominally low-impact self-XSS in a forum application can be turned into a real account-compromise issue by chaining it with web cache poisoning. The attack relies on g...

This hands-on CTF write-up models a Node.js supply-chain compromise against an application that consumes a private prisoner-db package from a local Verdaccio registry. The target stack includes a c...

Cisco Talos documents how attackers are abusing n8n’s public webhook feature to turn trusted n8n.cloud subdomains into phishing and malware-delivery infrastructure. In the observed campaigns, email...

The article describes an OAuth popup hijack caused by a hardcoded window.open(..., "addons-oauthWindow") target in an addons-linking flow, which lets an attacker pre-create a browsing context with ...

The article describes a path validation bypass in Bun applications caused by a normalization desync between WHATWG URL parsing and Node/POSIX path utilities. Bun preserves multiple leading slashes ...

This write-up describes an authorization bypass in LangSmith that chained two issues: an Agent API response exposed an internal X-Service-Key JWT, and a path normalization mismatch let requests rea...

The write-up describes an authenticated arbitrary file read in n8n’s Read/Write File node caused by a time-of-check/time-of-use race around symlink resolution. Even after a prior symlink-traversal ...

This research maps the URL-decoding pipeline in eight major frontend frameworks and shows that dynamic route parameters can be transformed into client-side path traversal primitives before develope...

Datadog describes extending its App and API Protection engine into Python AWS Lambda runtimes by embedding detection in the application process through its tracing library instead of relying only o...

Quarkslab reverse-engineers Ghidra’s BSIM pipeline and explains how the tool fingerprints functions for cross-compiler, cross-optimization, and cross-architecture similarity matching. The post show...

This primer maps the GitHub Actions threat model around workflows, runners, repository secrets, the built-in GITHUB_TOKEN, and third-party actions, showing how CI/CD automation becomes an attack su...

This write-up covers a watering-hole compromise of CPUID’s official download flow in which cpuid.com served legitimate signed CPU-Z packages bundled with a malicious CRYPTBASE.dll, turning the vend...

TrustedSec benchmarks self-hosted LLMs on offensive-security tasks by giving each model a minimal agent loop against OWASP Juice Shop: a system prompt, target-specific context, and just two tools f...

Cisco Talos’ 2025 review argues that state-sponsored operations from China, Russia, North Korea, and Iran differ in mission but converge on the same access patterns: rapid exploitation of newly dis...

Cisco Talos summarizes Microsoft’s April 2026 Patch Tuesday, which fixes 165 vulnerabilities across Windows, Office, .NET, Active Directory, SharePoint, and network-facing components, with eight ra...

SySS introduces wcfproxy, a Go-based interception proxy for analyzing Windows Communication Foundation services that use Net.TCP, where SOAP envelopes are encoded in WCF’s binary NBFX format instea...

This Patch Tuesday roundup covers 165 Microsoft fixes for April 2026, with emphasis on one actively exploited issue, one publicly disclosed issue, and several critical remote code execution bugs in...

Neil Madden analyzes Anthropic’s Mythos announcement as a vulnerability-finding system and argues the main technical question is not whether LLMs can hypothesize bugs, but whether they can reliably...

Cognisys shows a Citrix breakout in a locked-down Windows desktop where direct launches of cmd.exe and powershell.exe were blocked, but .bat files were still executable. That gap allowed the team t...

Binary Ninja 5.3 is a technically substantive tooling release centered on reverse-engineering workflows, adding new architecture APIs for standalone function-level lifting so plugins can analyze st...

This digest highlights two substantive AWS-focused research findings rather than simple news. The first covers Palo Alto Unit 42’s AgentCore work, where AWS Code Interpreter sandbox isolation can b...

OpenAnalysis documents an actively developed Rust/MSVC Windows malware family dubbed JitterDropper, with two variant lines that either decrypt an embedded payload from .rdata or fetch a 122-byte en...

The post explains how ESC8 in Active Directory Certificate Services can turn otherwise low-value authentication coercion bugs like PetitPotam or PrintNightmare into full domain compromise. If AD CS...

The write-up describes two admin-to-admin vulnerabilities in LibreNMS before 26.3.0: a stored XSS on the showconfig page and an authenticated RCE via configurable troubleshooting binaries. The XSS ...

delivr.to describes an integration with Push Security that correlates browser-layer detections from Push’s extension with specific payload links used in delivr.to web campaigns. The workflow polls ...

This article is a technical history of free-space optical communications that traces how infrared links evolved from niche laser networking and satellite experiments into short-range device interco...

This Hack The Box write-up walks through a Windows Server 2025 compromise that starts with valid MSSQL credentials and turns them into full domain admin. The attack abuses MSSQL login impersonation...

Datadog released an open source “AI-native” SAST engine that replaces purely rule-based scanning with a pipeline that first heuristically selects risky files, then pulls related functions/files to ...

This weekly roundup collects three technically relevant stories: a DOJ/FBI disruption of APT28 infrastructure that had compromised TP-Link routers, rewritten DNS settings, and returned forged recor...

The article analyzes a hardware USB keylogger that evades traditional antivirus and EDR by operating entirely outside the host OS as a USB HID passthrough device. Using the AirDrive Forensic Keylog...

Praetorian describes an indirect prompt injection technique against a multi-agent customer service system where a supervisor model screened only the user’s chat message, while the primary chat agen...

This Dayzerosec episode curates and discusses several substantive security-research topics rather than delivering a single advisory. The technical segment centers on low-level reverse engineering a...

Quarkslab researchers performed a full hardware teardown of a BYD Seal telematic unit (TCU) built around a Qualcomm MDM9628 (Snapdragon X5 LTE) SoC with a Micron MCP combining NAND flash and LPDRAM...

SentinelOne analyzes a pattern they call ‘edge decay’ — the systematic exploitation of perimeter devices (firewalls, VPN concentrators, load balancers) as primary intrusion vectors. Key mechanisms ...

Trail of Bits published a comprehensive C/C++ security testing checklist covering five areas: general bug classes (memory safety, integer errors, type confusion, compiler-introduced bugs), Linux us...

TrustedSec documents three Azure IAM permissions that enable privilege escalation: Microsoft.Authorization/roleAssignments/write (assign any role to any principal), Microsoft.Authorization/roleDefi...

CVE-2020-8562 is a TOCTOU vulnerability in the Kubernetes API server proxy that allows DNS rebinding attacks to bypass the isGlobalUnicast filter meant to block access to private IPs (e.g., 127.0.0...

Cisco Talos distills key findings from their 2025 Year in Review into actionable guidance for defenders and incident responders. Identity-based attacks accounted for 60% of Talos IR cases in 2024, ...

Cisco Talos highlights a ‘Platform-as-a-Proxy’ (PaaP) phishing technique where threat actors abuse legitimate SaaS notification pipelines (GitHub, Jira) to send phishing emails that pass SPF, DKIM,...

ProjectDiscovery details the prompt caching architecture behind Neo, their multi-agent security testing platform, which runs 26-step tasks with 40 tool calls and 20K+ token system prompts per agent...

Scanscope is an open-source Python tool that transforms nmap XML output into interactive 2D visualizations of large network port scan results. It models each host as a binary vector in a 2^17-dimen...

Crystal Mask is a Cobalt Strike sleepmask implementation built on the Crystal Palace/Tradecraft Garden framework, which separates evasion logic from payload capabilities via IAT hooking and PICO me...

Ingress NGINX reached EOL in March 2026, meaning future vulnerabilities (including cluster-wide impact bugs like IngressNightmare and CVE-2026-24512/CVE-2026-3288) will have no supported fixes. Thi...

Datadog’s threat modeling guide for GitHub CI/CD security maps attack surfaces across inputs (workflows, secrets, source code), identities (PATs, SSO, GitHub Apps), and risks (malicious code execut...

Datadog presents a MITRE ATT&CK-style threat matrix specifically mapped to CI/CD pipeline attack paths, covering tactics from Reconnaissance through Impact across SCM, CI, and CD trust boundari...

Unit 42 discloses ‘Agent God Mode’ — a multi-stage attack chain stemming from overly broad IAM roles auto-generated by the AWS Bedrock AgentCore starter toolkit. The default execution role applies ...

Cisco Talos documented UAT-10362, a threat actor conducting spear-phishing campaigns against Taiwanese NGOs and universities delivering ‘LucidRook’ — a novel stager that embeds a Lua interpreter an...

Red Canary’s 2026 Threat Detection Report overview covering the dual-edged role of AI in cybersecurity. On the offensive side, nation-state actors (Iran, China, North Korea) have used LLMs and MCP ...

A curated security newsletter highlighting several noteworthy items: Google’s GTIG published analysis of ‘DarkSword’, an iOS exploit chain; TeamPCP has been conducting active supply chain attacks (...

Talos’s 2025 Year in Review highlights the year’s most exploited vulnerabilities, with React2Shell (a 2025 RCE chain) spiking to the #1 attacked CVE in the final three weeks of the year, followed b...

Proprietary source code for virtually all major Yandex services was leaked via torrent on BreachForums in January 2023, with files dated to 24 February 2022. The leak covers backend code for Yandex...

Trail of Bits audited WhatsApp’s Private Inference system — a TEE-based architecture using AMD SEV-SNP and Nvidia confidential GPUs to process encrypted messages for AI features without exposing pl...

This article analyzes the practical attack surface that makes malicious ML models on HuggingFace a viable threat beyond just the model file itself. The primary vector highlighted is namespace/organ...

This research demonstrates how KServe — a Kubernetes-native ML model serving platform — can be abused to deploy malicious models as part of pre/post-exploitation in MLOps environments. The attack e...

Cisco Talos documents a technique dubbed Platform-as-a-Proxy (PaaP), where attackers abuse legitimate SaaS notification pipelines — specifically GitHub commit notifications and Jira Service Managem...

A CTF write-up for TryHackMe’s crackme8 challenge, walking through static and dynamic analysis of a 32-bit ELF binary. The binary has two sequential checks: argument count must equal 2, and the arg...

This article demonstrates remote user session enumeration on Windows hosts using undocumented WinStation APIs from winsta.dll — specifically WinStationOpenServerW, WinStationEnumerateW, and WinStat...

A Windows local privilege escalation technique abuses the Service Control Manager (SCM) by modifying its DACL via sc.exe sdset scmanager D:(A;;KA;;;WD), which grants Everyone full KEY_ALL_ACCESS ri...

This DEFCON31 talk write-up details supply chain attack techniques targeting ML pipelines via Hugging Face, analogous to Alex Birsan’s dependency confusion but applied to ML models. Three primary v...

This post documents the author’s journey through Windows Protected Process Light (PPL) internals with the goal of disabling PPL on lsass.exe via a kernel driver. It explains the PPL hierarchy — whe...

The Offsec ML Playbook v0.1 is a structured wiki of offensive ML TTPs organized into three domains: supply chain attacks, offensive ML techniques, and adversarial ML. Supply chain coverage includes...

MacNoise is a modular macOS telemetry noise generator written in Go, designed to help defenders audit EDR, SIEM, and firewall coverage by generating real system events: process spawns, network conn...

Unit 42 researchers discovered that AWS Bedrock AgentCore’s Code Interpreter sandbox mode — advertised as providing ‘complete isolation with no external network access’ — permits DNS resolution, en...

TrustedSec’s final installment in their detection foundation series focuses on correlating Windows event logs for detection engineering and incident response. The core technique uses the LogonID fi...

The article introduces the ‘adversary flywheel’ — a framework for red team operators to systematically collect attack telemetry and apply ML offensively, mirroring the data flywheel that defensive ...

A guide to LLM penetration testing covering the primary attack surfaces and vulnerability classes in AI-powered systems. Topics include direct and indirect prompt injection, training data and PII l...

Unit 42 presents two real-world Kubernetes attack case studies showing a 282% year-over-year rise in Kubernetes-targeted threat operations. In the first case, North Korean APT Slow Pisces (Lazarus/...

This HTB write-up for the hard-rated DarkZero box covers a two-forest Windows AD environment connected by a bidirectional cross-forest trust. The attack chain begins with MSSQL linked server abuse:...

Wiz traces a multi-account supply chain campaign (‘prt-scan’) exploiting the GitHub Actions pull_request_target trigger — a dangerous workflow event that runs with write permissions and secrets acc...