Threat Alerts

The rapid advancement of generative AI is creating a new challenge for organizations worldwide: the rise of fully fabricated job candidates. Threat actors can now leverage AI-generated profile photos, polished resumes optimized to bypass applicant tracking systems, and even deepfake video and voice technologies to create convincing identities that successfully navigate recruitment processes. As remote hiring becomes increasingly common, these synthetic candidates are becoming harder to distinguish from legitimate applicants, exposing weaknesses in traditional screening and interview practices.

As hiring increasingly relies on digital interactions, traditional recruitment controls are struggling to keep pace with evolving impersonation techniques. Standard resume reviews, interviews, and basic background checks are often insufficient to validate that a candidate’s identity, qualifications, and supporting documentation genuinely belong to the same individual. In response, organizations are strengthening identity verification procedures, credential validation processes, and interview-stage authentication measures to reduce the risk of fraudulent hires gaining access to corporate environments, sensitive information, and business operations.

Pink extortion group uses voice phishing to steal Microsoft 365 credentials and access cloud-hosted files, enabling rapid data theft and extortion that can cause reputational damage and business disruption.

Attacks begin with vishing calls impersonating internal IT to persuade employees to visit credential-harvesting sites. Captured credentials and active MFA sessions are abused to log into Microsoft 365 accounts, after which automated tooling and Graph API calls are used to find and download files from SharePoint and OneDrive. Compromised accounts are quickly leveraged to send internal extortion messages via email and collaboration channels to amplify pressure on victims.

Operators favor rapid data collection over persistence, reusing phishing infrastructure and customizing subdomains per target. Observed indicators include passkeyadd[.]com, passkeydeploy[.]com, deploypasskey[.]com, Microsoft.Graph.Client and python-requests user-agent strings, hosting tied to DDoS-Guard services, and several IPs used for phishing and proxying. Given the speed of exfiltration, organizations should assume credential compromise can lead to immediate data theft and act accordingly.

A highly sophisticated Chinese-speaking threat actor has expanded its operations beyond East Asia, increasingly targeting organizations across Europe, Africa and other global regions through a combination of malware delivery, credential phishing and financial fraud campaigns. Unlike most cybercriminal groups that focus on a single objective, this actor simultaneously pursues remote access, credential theft, phishing and payment card fraud, indicating a broad financially motivated strategy centered on data theft, access resale, persistent network access and monetization of compromised environments. The group relies heavily on localized social engineering lures themed around HR, payroll, taxation, invoicing and business communications, tailoring campaigns to specific regions while leveraging legitimate software, cloud-hosted infrastructure and trusted services to blend malicious activity with normal enterprise operations.

The actor has been observed deploying multiple malware families, including remote access trojans, loader frameworks and Python-based information stealers, while also conducting impersonation campaigns designed to move victims from email into messaging platforms where social engineering activity is less visible to traditional security controls. Analysis of malware source code, embedded strings, developer comments and reused constants suggests the group is likely leveraging large language models to accelerate development of new Python-based malware variants, enabling rapid expansion of its malware ecosystem. Its most advanced backdoor operates through a modular architecture, downloading core functionality and additional plugins from command-and-control infrastructure and shares characteristics with established Chinese command-and-control frameworks. Although currently assessed as financially motivated, the group’s advanced surveillance capabilities, evolving malware arsenal and expanding geographic targeting indicate a threat profile that could support both cybercriminal and espionage-oriented operations.

A newly cataloged vulnerability in a SolarWinds platform has been added to the U.S. government’s Known Exploited Vulnerabilities list.

Tracked as CVE-2026-28318 (CVSS Score 7.5), the flaw is an unauthenticated denial-of-service vulnerability, meaning attackers can exploit it without needing any valid credentials — raising the threat level considerably for organizations relying on the platform for file operations. The attack vector involves remote attacker sending a specially crafted HTTP POST request using a specific content encoding header, which causes the service to crash entirely. The impact is operational — file transfer services become unavailable to legitimate users, potentially disrupting business-critical workflows.

A publicly disclosed vulnerability in Visual Studio Code allows attackers to steal GitHub authentication tokens through a single user interaction, with no patch or CVE currently available. The flaw abuses VS Code’s sandboxed webview message-passing architecture, enabling malicious JavaScript running inside a webview to interact with the main editor and install a rogue extension. Once executed, the extension extracts the victim’s GitHub OAuth token and can use it to access every repository available to that user, rather than being restricted to a specific project or repository. The entire attack chain can be triggered by simply clicking a crafted link, requiring no elevated privileges or additional user interaction.

The malicious extension leverages simulated keypresses to automate installation, harvest the OAuth token and query GitHub APIs to enumerate accessible private repositories, creating a significant risk for developers and organizations relying on GitHub-based workflows. The vulnerability was publicly disclosed with only one hour of notice to the vendor, reportedly due to dissatisfaction with a previous vulnerability disclosure experience in which a security issue was fixed without acknowledgment of its security impact. The disclosure follows a broader trend of publicly released zero-days affecting Microsoft products, including recent privilege escalation vulnerabilities, a BitLocker bypass, and a flaw capable of disrupting antivirus definition updates. The incident highlights both the security risks associated with developer tooling and the growing friction between independent security researchers and enterprise vulnerability disclosure processes.

HTTP/2 Bomb is a newly disclosed denial-of-service (DoS) technique that combines HPACK header compression abuse with HTTP/2 flow-control manipulation to exhaust server memory and render services unavailable. The attack affects multiple widely deployed web servers, including nginx, Apache HTTP Server, Microsoft IIS, Envoy, and Cloudflare Pingora, with vulnerable behavior present in their default HTTP/2 configurations.

The attack exploits how HTTP/2 handles compressed headers and stalled streams. By sending specially crafted requests containing thousands of compressed header references and then preventing the server from completing responses, an attacker can force excessive memory allocations that remain pinned in memory. Researchers demonstrated amplification ratios of up to 5,700:1, allowing a single client on a standard broadband connection to consume tens of gigabytes of server memory within seconds. In testing, Apache HTTP Server and Envoy instances reached approximately 32 GB of memory consumption in under 20 seconds, causing significant service degradation and potential outages.

Following responsible disclosure, fixes have been released for some affected products. Apache addressed the issue under CVE-2026-49975, while nginx introduced new protections in version 1.29.8. At the time of publication, Microsoft IIS, Envoy, and Cloudflare Pingora had not yet released patches.