Threat Intelligence Brief - Friday, August 7, 2026

Operational threat reporting for defenders who need signal, not noise.

By DevSecOpsDad
Posted on August 7, 2026

Threat Radar

  • Microsoft identity infrastructure is under active, multi-vector attack — adversaries are exploiting Windows Hello for Business keys for stealthy Entra ID persistence while a separate AitM phishing campaign harvests Microsoft 365 credentials from finance and payroll teams.

  • AI coding agents introduced a zero-privilege supply chain attack path — a GitHub issue from an unprivileged account was sufficient to execute code on CI runners and exfiltrate workflow secrets in Anthropic’s Claude Code and Google’s Gemini CLI environments; OpenAI’s agent was also susceptible to run hijacking.

  • An Apache zero-day with no confirmed patch timeline is now public — discovered via AI-assisted analysis of 30,000 attack vectors, alongside novel HTTP desynchronization techniques that expand the attack surface for internet-facing web infrastructure.

  • Microsoft and Apple released critical patches this week — covering Azure, Entra, and SharePoint (Microsoft) and a high-severity authentication bypass (Apple); confirm patch status across enterprise environments before end of week.

  • 3.8 million health and insurance records were stolen from Unlimited Technology Systems — personal, medical, and health insurance data is confirmed exposed; healthcare and insurance organizations with vendor relationships should assess third-party data sharing exposure.



Immediate Action Required

  • Entra ID / Windows Hello for Business — Active Exploitation Confirmed: Malware operating within a signed-in Windows session can silently authenticate to Entra ID using Windows Hello for Business keys and register attacker-controlled devices for persistent cloud access. IAM and endpoint teams must audit Entra ID device registrations for anomalies, review conditional access policies, and confirm that privileged accounts carry controls beyond WHfB alone. (T1556, T1098)

  • Microsoft 365 AitM Phishing — Active Campaign Targeting Finance and Payroll: An active phishing campaign uses adversary-in-the-middle infrastructure to bypass MFA and hijack M365 sessions, with deliberate targeting of personnel in financial workflows. SOC teams should review sign-in logs for anomalous session tokens. Notify finance and payroll leadership directly. Prioritize FIDO2 deployment for high-value accounts. (T1598.003, T1187)

  • AI Coding Agent CI/CD Vulnerabilities — Demonstrated Exploitation: If your engineering organization uses Claude Code, Gemini CLI, or OpenAI coding agents integrated with GitHub CI/CD, treat pipeline secrets as potentially compromised. Audit workflow permissions, rotate exposed secrets, and restrict CI trigger permissions to trusted contributors immediately. (T1059, T1552.007)



High-Impact Developments

Malware Abuses Windows Hello for Business Keys for Persistent Entra ID Access

  • What happened: Researcher Dirk-jan Mollema demonstrated that malware executing within an active Windows session can silently leverage the user’s Windows Hello for Business cryptographic key to authenticate to Microsoft Entra ID — without requiring the user’s PIN or biometric. The attacker can then register a new device under their control, establishing durable cloud persistence that survives password resets.

  • Why it matters: This attack converts endpoint compromise into persistent cloud identity access. Password resets do not remediate it. The technique is stealthy, produces minimal user-visible indicators, and directly undermines the assumed security benefit of passwordless authentication.

  • Who should care: Identity and Access Management, Endpoint Security, Cloud Operations, Security Operations.

  • Recommended action: Audit Entra ID for unexpected device registrations, particularly those registered from endpoints with recent malware detections. Review conditional access policies to enforce device compliance checks. Confirm that privileged accounts require additional authentication controls beyond Windows Hello for Business alone.

  • Confidence: High — demonstrated by credible researcher with known exploitation confirmed.

  • Search metadata: T1556, T1098 — Windows Hello for Business, Microsoft Entra ID, Windows

Intelligence Context



Microsoft 365 AitM Phishing Campaign Targets Finance and Payroll Personnel

  • What happened: An active, widespread email-driven phishing campaign uses adversary-in-the-middle infrastructure to intercept Microsoft 365 authentication sessions, bypass MFA, and hijack accounts. Attackers are deliberately collecting email from personnel involved in financial workflows and payroll operations — a pattern consistent with business email compromise and payment fraud staging.

  • Why it matters: AitM techniques defeat standard MFA, making conventional phishing defenses insufficient. The deliberate targeting of finance and payroll personnel indicates the campaign is oriented toward financial fraud, not just credential harvesting. Compromised accounts give attackers internal email context to conduct convincing follow-on fraud.

  • Who should care: Finance, Payroll, Identity and Access Management, Security Operations.

  • Recommended action: Alert finance and payroll leadership immediately. Review M365 sign-in logs for token anomalies and unfamiliar session origins. Accelerate FIDO2/hardware key deployment for finance and payroll roles. Confirm that Conditional Access policies enforce compliant device requirements.

  • Confidence: High — active campaign with confirmed exploitation.

  • Search metadata: T1598.003, T1187 — Microsoft 365, AitM phishing, account hijacking

Intelligence Context



AI Coding Agents Expose CI/CD Pipeline Secrets to Zero-Privilege Attackers

  • What happened: Novee Security demonstrated that a GitHub issue submitted by an account with no repository privileges was sufficient to trigger code execution on CI runners used by Anthropic (Claude Code) and Google (Gemini CLI), and to hijack agent runs on OpenAI’s pipeline. Workflow secrets — including API keys, tokens, and credentials stored in CI environments — were accessible as a result.

  • Why it matters: The attack requires no privileges: any external actor who can open a GitHub issue can potentially compromise the CI pipeline and exfiltrate secrets. This was demonstrated against the vendors’ own repositories. Organizations that have adopted AI coding agents without tightly scoping pipeline permissions face the same vulnerability class.

  • Who should care: Software Engineering, DevOps, Security Operations, Third-Party Risk.

  • Recommended action: Audit GitHub Actions workflow trigger permissions immediately — restrict pull_request_target and issue-triggered workflows to trusted contributors. Rotate secrets stored in CI environments for affected agent integrations. Confirm whether AI coding agent integrations require the level of CI/CD access currently granted.

  • Confidence: High — demonstrated exploitation against vendor-operated repositories confirmed.

  • Search metadata: T1059, T1552.007 — Claude Code, Gemini CLI, GitHub, CI/CD, secret exposure

Intelligence Context



Apache Zero-Day and Novel HTTP Desync Techniques Disclosed

  • What happened: PortSwigger’s AI-assisted HTTP Terminator research tool, developed by James Kettle, analyzed 30,000 candidate attack vectors and identified new HTTP desynchronization techniques. A separate human-guided research cascade uncovered an Apache zero-day vulnerability. No patch availability timeline has been confirmed.

  • Why it matters: HTTP desynchronization attacks enable request smuggling, cache poisoning, and unauthorized access to backend systems. An unpatched Apache zero-day exposes internet-facing infrastructure before mitigations are broadly available. AI-assisted discovery of these techniques signals that the HTTP attack surface is expanding faster than traditional research methods reveal.

  • Who should care: Infrastructure teams, web application owners, Security Operations.

  • Recommended action: Inventory internet-facing Apache deployments. Monitor Apache security advisories for patch release. Assess whether WAF rules or reverse proxy configurations can provide interim mitigation for HTTP desync attack patterns pending patch availability.

  • Confidence: Medium — vulnerability confirmed by credible researcher; exploitation status unknown.

  • Search metadata: T1190 — Apache, HTTP desynchronization, zero-day

Intelligence Context



Microsoft and Apple Release Critical Security Patches

  • What happened: Microsoft released critical patches covering Azure, Entra, and SharePoint. Apple patched a high-severity authentication bypass vulnerability. Exploitation status for these specific vulnerabilities is not confirmed, but the affected products are high-value targets.

  • Why it matters: Azure, Entra, and SharePoint are core enterprise infrastructure. Given the active identity threats against Microsoft environments in this brief, unpatched vulnerabilities in these products represent compounding risk. Apple’s authentication bypass warrants prompt attention for organizations with significant Apple device fleets.

  • Who should care: IT Operations, Cloud Operations, Identity and Access Management, Security Operations.

  • Recommended action: Confirm patch deployment status for Azure, Entra, and SharePoint updates across the enterprise. Validate Apple patch rollout through MDM. Prioritize Entra patches given the concurrent active identity threats in this brief.

  • Confidence: High — vendor-confirmed patches; exploitation status unknown.

  • Search metadata: Azure, Entra, SharePoint, Apple, authentication bypass

Intelligence Context



Monitor Only



Analyst Observation

Today’s brief reflects a Microsoft-heavy threat environment that is not coincidental — identity infrastructure, cloud access, and collaboration platforms are the primary attack surface for both opportunistic and targeted actors right now. The Windows Hello for Business persistence technique is particularly concerning because it undermines a control many organizations deployed specifically to reduce credential risk; IAM teams that have treated passwordless authentication as a solved problem need to revisit that assumption. The AI coding agent CI/CD findings deserve more attention than they will likely receive: the attack requires no privileges, the blast radius includes pipeline secrets and potentially source code, and the pattern generalizes well beyond the three vendors tested. Organizations that have adopted AI coding agents without reviewing CI/CD trigger permissions and secret scoping have an unacknowledged exposure that should be closed this week.





Generated by DevSecOpsDadAttack cyber threat intelligence.

Tags:  Cyber-Security-News  T1598.003  T1187  T1190  T1499  T1557.002  T1040  T1556  T1098  T1059  T1552.007  Microsoft-365
DevSecOpsDadAttack!
Share: X (Twitter) LinkedIn