CVE-2026-54200 is a path filtering bypass in TeamDavid's @@attach mechanism. The filter checks whether requested file paths contain "David config" or "user folder" strings to prevent access to sensitive directories. This protection can be defeated by appending an NTFS Alternate Data Stream (ADS) syntax—specifically, a colon followed by a stream name—to the path. For example: "user:secret.txt" resolves to the ADS "secret.txt" on the user folder, bypassing the string-based filter while the OS still resolves the underlying resource.
This works because ADS are first-class filesystem objects on NTFS that resolve during file access but don't appear in directory listings or normal path string operations. A filter checking whether a path string contains "user folder" will pass "user folder:something" because the literal substring isn't present—yet the operating system resolves this to an actual file in the user folder's alternate stream namespace. Windows APIs don't expose ADS through standard enumeration, so the filter was checking a sanitized view of the filesystem that doesn't match what actually gets accessed.
The @@attach feature itself is the architectural problem. It grants authenticated users the ability to read arbitrary files on the system by name, with no apparent scoping to user-controlled directories. This is a privilege model that made sense for internal operations but became dangerous once exposed as an attachable resource. Combined with the ADS bypass, any file readable by the service account becomes exfiltratable.
What to check: First, verify whether TeamDavid installations are on Windows and whether Rollout 524 (or subsequent patches) has been applied. If unpatched, the risk is immediate—any authenticated user with access to @@attach can read any file the service account can access, including credential stores, private keys, and configuration files containing secrets. Second, review whether @@attach has been scoped to a restricted allowlist of directories (user-writable temp folders only) rather than operating with broad filesystem access. Third, implement application-layer controls that validate not just path strings but the actual resolved resources—checking the final NTFS object after all parsing and stream resolution, not just the input string.
The deeper issue is that string-based path filtering is fundamentally inadequate on Windows. ADS is one variant; short-name aliases, Unicode normalization edge cases, and device path conventions (\.\, \?) represent the same class of bypass. A correct fix normalizes and validates the final resolved path, not the input. If the vendor's remediation is limited to blocking ADS syntax specifically, expect another bypass in this category within months.