Happy Friday!
This week, two previous IAI deliverables, CVE-2026-5027 (Langflow) and CVE-2026-39808 (FortiSandbox), were added to the VulnCheck KEV after exploitation was observed in our canary network. Additionally, our research organization published two excellent blogs that may be of interest: Copy Fail and Its Descendants and From Canary Intelligence to C2: Mapping an Attack Fleet with Target Intelligence.
Otherwise, here are your regularly scheduled Initial Access Intelligence weekly release notes:
The team developed an unauthenticated remote code execution chain consisting of CVE-2026-34908, CVE-2026-34909, and CVE-2026-34910 affecting Ubiquiti UniFi OS. The chain was initially disclosed by Bishop Fox and augmented with an additional privilege escalation, resulting in code execution as root. CVE-2026-34908 and CVE-2026-34910 were added to the VulnCheck KEV on June 9, 2026, following observed exploitation by XServus. Additionally, our Censys query identifies approximately 10,000 internet-exposed UniFi OS Server instances, suggesting exploitation activity may accelerate.
Our coverage includes the chained exploit with root-escalation and unprivileged-shell payload options, encrypted and unencrypted PCAPs, Suricata and Snort rules, and ASM queries for Shodan, Censys, FOFA, ZoomEye, and GreyNoise.
At customer request, the team developed an exploit for CVE-2021-21974, affecting VMware ESXi, that allows a remote, unauthenticated attacker to gain full root access on the hypervisor. The vulnerability, added to the VulnCheck KEV in February 2023, is known to be used by ransomware groups ESXiArgs, RansomExx2, and Royal Ransom. Additionally, Cosmic Wolf (aka Sea Turtle) has been known to exploit this vulnerability.
VulnCheck continues to observe a concerning number of internet-facing ESXi servers, with more than 15,000 reportedly running the 6.7.0 branch targeted by our exploit. Alongside the exploit, we also shipped a version scanner, PCAPs, and Suricata and Snort detection rules.
Added this week was coverage for an authenticated command injection vulnerability in the D-Link DIR-823x router affecting the /goform/diag_ping endpoint. Although we have not observed exploitation of this specific vulnerability, D-Link routers remain a common target for threat actors, and the DIR-823x series has been targeted recently. CVE-2025-29635, another authenticated command injection vulnerability, was observed in a Mirai botnet campaign in March 2026, according to an Akamai report. CVE-2025-29635 was also added to both the VulnCheck and CISA KEV lists. More broadly, Rondodox campaigns have been linked to exploitation of D-Link IP cameras through CVE-2013-1599 and D-Link NAS devices via CVE-2024-10914.
Coverage includes an exploit, a PCAP, network signatures, and ASM queries.
The team developed an exploit for CVE-2026-34159, an unauthenticated remote code execution vulnerability in the RPC server (rpc-server) shipped with llama.cpp, the open-source C/C++ LLM inference engine. Although not yet known to be exploited in the wild, multiple public exploits exist, including two well-written blog posts: CVE-2026-34159: The Deserializer Three CVEs Have Not Patched and CVE-2026-34159: Exploiting llama.cpp’s RPC Server - From Null Buffer to RCE Against PIE + Full RELRO + NX, increasing the likelihood of future exploitation.
Our exploit comes with a target Docker container, a packet capture, and Suricata and Snort rules.
The team developed an exploit for CVE-2026-8181, an unauthenticated remote code execution vulnerability in Burst Statistics, a privacy-friendly WordPress analytics plugin with more than 200,000 active installations. This vulnerability first caught the team's attention when a public exploit was pushed to GitHub alongside a list of targets using Burst Statistics. While we don't currently have proof of exploitation, we think... uh... it's probably not a coincidence.
Our exploit comes with a version scanner, a target Docker container, encrypted and unencrypted PCAPs, and Suricata and Snort rules.
The team added ASM queries for CVE-2026-50751, affecting Check Point systems and allowing an attacker to bypass the IKE handshake and directly establish a VPN connection. Our Censys queries currently identify approximately 100,000 potentially exposed devices. The vulnerability was added to the VulnCheck KEV on June 8, 2026, and has been linked to at least one Qilin ransomware affiliate. The team is prioritizing additional coverage for this vulnerability next week.
The team developed an exploit for CVE-2026-25874, an unauthenticated remote code execution vulnerability in Hugging Face's LeRobot async inference policy server. The LeRobot GitHub repository has more than 24,000 stars, and this vulnerability has been featured in multiple blogs, including one by our own Valentin Lobstein.
Our exploit comes with a target Docker container, a packet capture, and Suricata and Snort rules.
The team developed an exploit for CVE-2026-42589, an unauthenticated remote code execution vulnerability in Gotenberg, a Docker-powered stateless API for converting HTML, Markdown, and Office documents into PDFs. Gotenberg is intended to be deployed as an internal document-conversion microservice, but unauthenticated internet-exposed instances remain accessible; our Shodan query identifies them. This vulnerability also has a public Nuclei template, increasing the likelihood of future exploitation.
Our exploit comes with a version scanner, a target Docker container, encrypted and unencrypted PCAPs, and Sigma, Suricata, and Snort rules.
The team developed an exploit chain using CVE-2024-46507 and CVE-2024-46508 to achieve unauthenticated remote code execution on Yeti Platform. Both vulnerabilities were originally discovered by Rhino Security Labs, and CVE-2024-46507 was added to the VulnCheck KEV in June 2025.
Included with the exploit are network rules, PCAPs, and ASM queries.