OpenSSL CVE-2026-54874

Exploitation requires that the target be running an OpenSSL DTLS server endpoint - TLS-only servers over TCP are not affected. The CVSS vector (AV:N/AC:L/PR:N/UI:N) confirms remote unauthenticated exploitation with no special configuration: any DTLS server using a vulnerable OpenSSL version is exposed by default simply by accepting inbound DTLS connections. No non-default configuration flag or feature toggle is required to trigger the vulnerable buffering path. Key factors that limit exploitation impact: the attack is scoped exclusively to DTLS (UDP-based), not TLS over TCP; per-connection memory consumption is bounded at approximately 1.7 MB, so total impact scales linearly with the server's concurrent DTLS connection limit; applications that already cap simultaneous DTLS associations proportionally limit worst-case memory loss; and the vulnerability provides no code execution, data exfiltration, or integrity compromise - availability is the sole impact dimension.

The official CVSS 7.5 vector (AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H) accurately reflects the network-reachable, unauthenticated nature of the attack and high availability impact, yet OpenSSL itself assesses severity as Low because the per-connection memory footprint is bounded at approximately 1.7 MB and any application-level ceiling on concurrent DTLS associations directly caps total exposure. SSVC corroborates this framing: exploitation status is confirmed none, the attack is automatable (an adversary can script connection floods), and technical impact is only partial. No EPSS score was provided in the input, so exploitation probability cannot be quantified numerically. The absence of a CISA KEV listing and the lack of any identified public exploit indicate no known active abuse at time of analysis. The amplification factor of ~1,200× is noteworthy - a relatively modest number of concurrent handshakes can exhaust server memory - but this scales with the server's connection limit, meaning operators who already impose strict concurrency caps have inherently reduced exposure. Impact is restricted entirely to availability; there is no confidentiality or integrity risk. Overall this is a maintenance-level patch priority: real and automatable, but bounded, not actively exploited, and remediable through standard upgrade cycles.

An unauthenticated remote attacker opens a large number of concurrent DTLS handshakes with a target server and, during each handshake, transmits a stream of small forged UDP datagrams bearing records that claim to belong to the next handshake epoch. Each such record causes OpenSSL to buffer a full 16 KB read buffer; with up to 100 records buffered per connection, each association consumes approximately 1.7 MB of server memory from minimal transmitted data. Multiplied across hundreds or thousands of concurrent connections (within the server's accept limit), the attacker drives the server toward memory exhaustion and denial of service. No public exploit code has been identified at time of analysis.