Eclipse Ankaios
Monthly
Let me analyze this CVE. CVE-2026-86836: Eclipse Ankaios 0.1.0 through 1.0.2. Agent creates workload files and Control Interface FIFOs under predictable path derived from agent name and hash of workload's runtime config. If directory/FIFO already exists at that path when agent (re)starts, agent reuses based only on existence/file-type check, without validating owner or permissions. Local unprivileged user with write access to same base directory (default `$TMPDIR/ankaios`, e.g. shared /tmp) can pre-create path hierarchy including two CI FIFOs before agent starts. Agent then treats attacker-owned FIFOs as legit Control Interface. Attacker can complete handshake and issue requests using workload's controlInterfaceAccess permissions, allowing impersonation and unauthorized read/modify of cluster's desired state. CVSS 8.4, CVSS:4.0/AV:L/AC:L/AT:P/PR:L/UI:N/VC:H/VI:H/VA:N/SC:N/SI:N/SA:H — wait, AT:P (attack requirements: present). Interesting. SA:H (subsequent availability high) — hmm, that's odd, maybe SC/SI/SA... Actually SA:H means subsequent system availability high. Whatever. CWE-379: Creation of Temporary File in Directory with Insecure Permissions. Fix: v1.0.3, PR #799. SSVC: Exploitation none, Automatable no, Technical Impact total. Tags: Authentication Bypass, Eclipse Ankaios, Eclipse Foundation. Prevalence: Eclipse Ankaios is a fairly niche open-source container/workload orchestration for embedded/automotive edge. Medium-low. I'd say "low" — it's niche, automotive/embedded focused, relatively new (0.x-1.x). Let's say low. Product name: "Eclipse Ankaios" — 2 words. Good. Now write fields. prevalence_basis: "niche embedded/automotive workload orchestrator, limited deployment" — max 12 words. Assessed CVSS 3.1: AV:L/AC:L/PR:L/UI:N/S:C? Hmm, does scope change? The attacker impersonates a workload and can read/modify cluster desired state — arguably scope change (S:C) because the vulnerable component (agent) impacts the cluster control plane. But let's ke
Incorrect authorization in Eclipse Ankaios v0.5.1-v1.0.1 allows an authenticated workload to bypass its scoped access restrictions on the agent-side Control Interface. When the cluster is configured with multi-segment allow rules whose first path segment is a wildcard (e.g., `*.workloads.some_workload`), submitting a CompleteStateRequest or UpdateStateRequest with an empty field mask is incorrectly evaluated as matching the scoped rule, granting the workload full read or write access beyond its authorized state subtree. No public exploit is identified at time of analysis; a patched release (v1.0.2) is available.
Uncontrolled memory allocation in Eclipse Ankaios 0.1.0-1.0.1 allows a workload with granted Control Interface access to abort the agent process by declaring an arbitrarily large protobuf message length over the Control Interface FIFO, causing unbounded heap allocation and an OOM-triggered process crash. The impact is complete loss of orchestration services for all workloads managed by the affected agent instance. No public exploit has been identified and no active exploitation is confirmed; patch v1.0.2 is available.
Let me analyze this CVE. CVE-2026-86836: Eclipse Ankaios 0.1.0 through 1.0.2. Agent creates workload files and Control Interface FIFOs under predictable path derived from agent name and hash of workload's runtime config. If directory/FIFO already exists at that path when agent (re)starts, agent reuses based only on existence/file-type check, without validating owner or permissions. Local unprivileged user with write access to same base directory (default `$TMPDIR/ankaios`, e.g. shared /tmp) can pre-create path hierarchy including two CI FIFOs before agent starts. Agent then treats attacker-owned FIFOs as legit Control Interface. Attacker can complete handshake and issue requests using workload's controlInterfaceAccess permissions, allowing impersonation and unauthorized read/modify of cluster's desired state. CVSS 8.4, CVSS:4.0/AV:L/AC:L/AT:P/PR:L/UI:N/VC:H/VI:H/VA:N/SC:N/SI:N/SA:H — wait, AT:P (attack requirements: present). Interesting. SA:H (subsequent availability high) — hmm, that's odd, maybe SC/SI/SA... Actually SA:H means subsequent system availability high. Whatever. CWE-379: Creation of Temporary File in Directory with Insecure Permissions. Fix: v1.0.3, PR #799. SSVC: Exploitation none, Automatable no, Technical Impact total. Tags: Authentication Bypass, Eclipse Ankaios, Eclipse Foundation. Prevalence: Eclipse Ankaios is a fairly niche open-source container/workload orchestration for embedded/automotive edge. Medium-low. I'd say "low" — it's niche, automotive/embedded focused, relatively new (0.x-1.x). Let's say low. Product name: "Eclipse Ankaios" — 2 words. Good. Now write fields. prevalence_basis: "niche embedded/automotive workload orchestrator, limited deployment" — max 12 words. Assessed CVSS 3.1: AV:L/AC:L/PR:L/UI:N/S:C? Hmm, does scope change? The attacker impersonates a workload and can read/modify cluster desired state — arguably scope change (S:C) because the vulnerable component (agent) impacts the cluster control plane. But let's ke
Incorrect authorization in Eclipse Ankaios v0.5.1-v1.0.1 allows an authenticated workload to bypass its scoped access restrictions on the agent-side Control Interface. When the cluster is configured with multi-segment allow rules whose first path segment is a wildcard (e.g., `*.workloads.some_workload`), submitting a CompleteStateRequest or UpdateStateRequest with an empty field mask is incorrectly evaluated as matching the scoped rule, granting the workload full read or write access beyond its authorized state subtree. No public exploit is identified at time of analysis; a patched release (v1.0.2) is available.
Uncontrolled memory allocation in Eclipse Ankaios 0.1.0-1.0.1 allows a workload with granted Control Interface access to abort the agent process by declaring an arbitrarily large protobuf message length over the Control Interface FIFO, causing unbounded heap allocation and an OOM-triggered process crash. The impact is complete loss of orchestration services for all workloads managed by the affected agent instance. No public exploit has been identified and no active exploitation is confirmed; patch v1.0.2 is available.