Sangoma Switchvox Vulnerabilities Exploited in the Wild
CVE-2026-9586 is a critical unauthenticated SQL injection vulnerability (CVSS 9.3) in Sangoma Switchvox VoIP management that allows remote code execution via unsanitized XML input in PostgreSQL queries Threat actors are actively exploiting CVE-2026-9586 in the wild, with Horizon3 sharing indicators of compromise to help organizations detect intrusions CISA added five additional vulnerabilities to its Known Exploited Vulnerabilities catalog, including CVE-2026-48710 (HTTP smuggling in Starlette),
Analysis
TL;DR
- CVE-2026-9586 is a critical unauthenticated SQL injection vulnerability (CVSS 9.3) in Sangoma Switchvox VoIP management that allows remote code execution via unsanitized XML input in PostgreSQL queries
- Threat actors are actively exploiting CVE-2026-9586 in the wild, with Horizon3 sharing indicators of compromise to help organizations detect intrusions
- CISA added five additional vulnerabilities to its Known Exploited Vulnerabilities catalog, including CVE-2026-48710 (HTTP smuggling in Starlette), CVE-2026-49869 (command injection in Kestra), and CVE-2026-59822 (authentication bypass in LiteLLM)
- CISA mandates federal agencies patch the critical flaws within three days, with Kestra and Starlette vulnerabilities allowed a two-week window under BOD 26-04
Why It Matters
This highlights the accelerating pace of real-world exploitation of critical vulnerabilities in widely deployed enterprise and open-source infrastructure, with SQL injection and authentication bypass flaws remaining persistent attack vectors. The inclusion of LiteLLM and Kestra vulnerabilities underscores the growing attack surface created by AI/ML tooling and orchestration platforms that organizations are rapidly adopting without adequate security hardening.
Technical Details
- CVE-2026-9586 (Sangoma Switchvox): Unauthenticated SQL injection in an XML-processing endpoint where the user-controlled
PhoneIPparameter is concatenated directly into PostgreSQL queries without sanitization or parameterization, enabling arbitrary SQL execution and remote code execution from a single crafted request - CVE-2026-48710 (Starlette): HTTP request/response smuggling vulnerability in the lightweight ASGI framework, exploited since its public disclosure in May
- CVE-2026-49869 (Kestra): Critical-severity command injection defect in the open-source orchestration platform, flagged as actively exploited by Microsoft
- CVE-2026-59822 (LiteLLM): High-severity authentication bypass vulnerability detected in the wild by Wiz honeypots
- CISA's KEV catalog additions reflect a mix of enterprise software (Sangoma, JFrog Artifactory, SonicWall) and open-source AI infrastructure components (LiteLLM, Kestra, Starlette)
Industry Insight
Organizations using AI/ML infrastructure tools like LiteLLM and Kestra should prioritize immediate patching and review access logs for signs of compromise, as these emerging platforms are becoming attractive targets for threat actors. The rapid exploitation cycle—where vulnerabilities move from disclosure to active exploitation within days—demands that security teams adopt continuous vulnerability monitoring and automated patch management rather than relying on periodic review cycles. The CISA BOD 26-04 directives signal increasing regulatory pressure on federal agencies and their contractors, which will likely cascade into stricter security requirements across government supply chains and private sector vendors.
Disclaimer: The above content is generated by AI and is for reference only.