New coverage for OpenJDK, NGINX, Check Point, Windows, JetBrains TeamCity, Cisco NX-OS, N-able, and many more.

Happy Friday! The team added 250+ new exploits this week, including full RCE exploits for Java4Shell and NGINX2root, neither of which is easily or universally exploitable. Some selected highlights from the week are below, with full details available in the Initial Access Intelligence API.

CVE-2026-114938: Oracle OpenJDK AVX-512 Base64 Decode Intrinsic Heap Buffer Overflow RCE

The team developed several exploit variants for CVE-2026-114938, aka "Java4Shell," which was disclosed earlier this week amid a truly impressive amount of FUD. The vulnerability arises from an integer underflow that leads to an out-of-bounds heap write in the HotSpot AVX-512 Base64 decode intrinsic in OpenJDK 17. Exploitation is hardware-specific and requires non-trivial heap grooming combined with a target-specific memory disclosure path to leak addresses and bypass ASLR. Our team developed a full RCE exploit the same day the vulnerability was disclosed (some weeks it seems like developing novel memory leaks is all they do), along with PCAPs and Suricata, Snort, Sigma, and YARA rules. We have also weaponized the vulnerability in Terrakube and Jenkins for unauthenticated code execution. For more background, read our emerging threat blog.

CVE-2026-123633: F5 NGINX HTTP/2 Reverse Proxy Connection Reuse Use-After-Free Unauthenticated Remote Code Execution

The team wrote an exploit for CVE-2026-123633, a use-after-free in NGINX that yields unauthenticated code execution as the worker user. We covered it because HTTP/2 in front of a proxied backend is one of the most common NGINX deployments, and public details stop short of a working exploit. Ours lands a shell in about 20 seconds, 10 of 10 times against our target; it does need a single-worker server that can connect back to the attacker. It is not yet on any KEV list. VulnCheck Target Intelligence reports roughly 151,000 internet-facing NGINX 1.31.8 servers (the build our exploit targets). Our exploit ships with a version scanner, a Docker target, and ASM queries; there are no network rules because the traffic exists only inside TLS.

CVE-2026-123007: F5 NGINX HTTP/2 Connection Send Window Underflow Unauthenticated Remote Code Execution

The team also added an exploit for CVE-2026-123007, an HTTP/2 flow-control accounting flaw in NGINX that yields unauthenticated code execution as the worker user. It was disclosed on November 18 and credited to at least four different research firms. NGINX fronts a large share of the internet, and HTTP/2 over TLS is a stock configuration. The bar for exploitation is high, however — roughly 119 GiB over a direct, unproxied connection to a freshly started worker, meaning mass exploitation is extremely unlikely. It is not yet on any KEV list. VulnCheck Target Intelligence reports roughly 147,600 internet-facing NGINX 1.31.8 servers. Our exploit ships with a version scanner, a Docker target, and ASM queries.

CVE-2026-32157: Microsoft Remote Desktop Client RDPDR Reassembly Use-After-Free RCE

The team developed an exploit for CVE-2026-32157, a use-after-free in the Microsoft Remote Desktop client (msrdc) and the Microsoft Terminal Services client (mstsc) that lets a malicious RDP server run code on any vulnerable client that connects to it with default settings. Microsoft rates it "Exploitation Less Likely," but our exploit defeats ASLR and CFG to reliably land a shell across multiple client builds. Rogue RDP servers are a proven lure: Midnight Blizzard phished targets with .rdp files pointing at actor-controlled servers. No in-the-wild exploitation has been reported. Our exploit ships with PCAPs and Snort and Suricata rules.

CVE-2026-85103: Check Point Security Gateway IKE Certificate Decoding Heap Overflow

The team added an exploit for CVE-2026-85103, a pre-authentication heap overflow in Check Point Security Gateway's VPN certificate decoding, disclosed alongside CVE-2026-85102 (which we covered earlier this year). No technical details are available on the vulnerability. Our team turned a one-byte overflow into an unauthenticated root shell on gateways with Remote Access VPN enabled, typically within a minute. The bug has been on VulnCheck KEV since September 25, 2026, based on a report from EUVD. It is not yet on CISA KEV. VulnCheck Target Intelligence sees roughly 26,000 Check Point hosts exposing the topology port that Remote Access clients use. Our exploit ships with a PCAP and ASM queries. The attack travels inside the encrypted IKE exchange, so network rules are not included.

CVE-2026-76465: Cisco NX-OS MPLS OAM Echo-Request Unauthenticated Remote Code Execution

The team developed an exploit for CVE-2026-76465, a heap corruption vulnerability affecting the MPLS OAM feature of Cisco NX-OS software running on Cisco 3000 and 9000 series switches. Successful exploitation of this vulnerability, achieved by sending specially crafted MPLS OAM echo requests, grants arbitrary code execution as the root user on the switch. It is not known to be exploited in the wild. Our ASM queries currently identify approximately 3,500 internet-exposed devices running NX-OS software. Our exploit ships with a PCAP, network signatures, and ASM queries.

CVE-2026-133089: JetBrains TeamCity TFS Helper Unauthenticated Command Server RCE

The team wrote an exploit for CVE-2026-133089, the one unauthenticated RCE candidate in JetBrains' large November 2026 TeamCity release. TeamCity's out-of-process TFS helper exposes a command server with no authentication; on builds before 2026.1.5 it also listens on every interface. The CVE record has no technical detail, and there are no public PoCs; our research proved full unauthenticated RCE, writing a webshell into the web root and running commands as the TeamCity service account. A compromised TeamCity server generally yields source code, credentials, and deployment access to everything downstream. No in-the-wild exploitation has been reported. Our exploit ships with a version scanner, a Docker target, PCAPs, Snort and Suricata rules, a YARA rule, and ASM queries.

CVE-2026-125883: N-able N-central ImportServiceFromFile XXE and Loopback ServerUI SSRF Unauthenticated RCE

Finally, the team added an exploit for CVE-2026-125883, a critical XML external entity vulnerability in N-able N-central. N-able's advisory describes it only as "XML Schema Poisoning"; our team turned it into unauthenticated RCE as the N-central service user, with the only precondition being the ID of any agent or probe registered with the server. The vulnerability was added to VulnCheck KEV on November 11, 2026, the day after it was published; it is not on CISA KEV at time of writing, but public exploit code for this and a number of other issues was published on November 13. N-central is RMM software with reach into every endpoint it manages, and a proven target, with a number of previous bugs that have been exploited in the wild. VulnCheck Target Intelligence identifies roughly 2,200 internet-exposed N-central instances. Our exploit ships with PCAPs, Snort and Suricata rules, and ASM queries.