Halil, the chain has three distinct boundaries. Initial access: CVE-2026-109618 and CVE-2026-109619 are unauthenticated PHP2Shell flaws exposed through vulnerable Mobile Access web interfaces. Privilege escalation: CVE-2026-123039 is a separate local flaw in Gaia’s DynamicCLI service; the daemon runs as root and exposes /tmp/dynamic_socket, allowing the already-established portal context to reach root. Seeing only exploit traffic for the first pair is attempted access; root commands or abnormal access to that socket establish the escalation stage. [1][2]
After escalation, Check Point reports PHP web shells, the persistent TSGA backdoor installed as slot_CPconfigdaemon, and its embedded TSProxy networking component. TSGA/TSProxy are linked technically to the previously reported PhiliKit-associated toolset, but that is malware-lineage evidence—not sufficient actor attribution. Hunt for unexpected PHP files or portal-worker child processes, slot_CPconfigdaemon, root-level filesystem/process changes, abnormal DynamicCLI socket access, suspicious gateway egress or proxy tunnels, and configuration access. The reported source IP 130.94.19.204 is useful corroboration, not proof by itself. [1][3]
The lateral stage is equally decisive: the investigated intrusion used SSH from the gateway to VMware ESXi, followed by destroy_v6, recursive deletion/zeroing, and DREAM_ANNEXATION.txt. The destructive activity belongs to the initial wave targeting Israeli organizations beginning around October 30. Check Point describes the broader November 2 international wave as affecting multiple countries and sectors without observed destructive activity. That distinction matters: international compromise should not automatically be labeled a wiper incident, while the political note should not be treated as reliable attribution. [1][3]
My response threshold is firm: a web shell, slot_CPconfigdaemon/TSGA, TSProxy-like tunneling, root execution through DynamicCLI, unexplained configuration access, or gateway-originated SSH into ESXi is enough to declare compromise—even after patching. Isolate the gateway, preserve disk and volatile evidence, rotate credentials and trust material accessible through it, investigate downstream systems, and rebuild from a trusted image. Patching alone closes the entry point; it does not establish that persistence or stolen credentials are absent.