IBM i CVE-2026-18251

Two conditions must both hold. First, the targeted IBM i system must have a WebSocket-enabled service reachable by the victim's browser - this is a non-default or optional service configuration on IBM i, not something present on a bare OS install, and the specific service and port were not disclosed in the available data. Second, the victim must be authenticated to that IBM i service and must be induced to load attacker-controlled web content in the same browser, because the CVSS vector is PR:N with UI:R - the attacker needs no credentials but does need the victim's cooperation. Exploitation is further limited by S:U and C:L/I:N/A:N: only confidentiality of a limited data set is exposed, no code execution, no data modification, and no denial of service. An unpublished or unauthenticated WebSocket endpoint would not be exploitable in this way.

Every signal points to a low-priority issue. CVSS 3.1 scores 4.3 with the notable metric UI:R - the attacker cannot trigger this alone and must convince a user to load malicious content in a browser while authenticated to IBM i; the impact is confined to C:L (limited confidentiality), with no integrity or availability effect (I:N/A:N). CISA's SSVC assessment corroborates this: Exploitation 'none', Automatable 'no', Technical Impact 'partial'. No EPSS value was supplied, no CISA KEV listing exists, and no public proof-of-concept was reported. The CVSS vector and the CWE are consistent with each other (network-reachable WebSocket endpoint, no privileges, but mandatory user interaction), so there is no disagreement to flag - however, because EPSS data and a numeric CVSS 4.0 score were absent from the input, the residual uncertainty is about how attractive this is to opportunistic scanning rather than about severity. Treat as a routine patch-cycle item, not an incident.

An attacker hosts a page containing JavaScript that opens a WebSocket to a known IBM i host. When an authenticated IBM i user visits that page, the browser completes the handshake without IBM i rejecting the foreign Origin, and the attacker reads whatever limited data the socket returns - session metadata or low-sensitivity application responses. No proof-of-concept is publicly identified, and exploitation depends entirely on the victim being logged in to IBM i and browsing to attacker-controlled content.