Zephyr RTOS CVE-2026-0849

Exploitation requires physical or proximate access to the I2C or SPI bus connecting the ATAES132A chip to the host microcontroller - either via a direct hardware bus tap, a logic analyzer with injection capability, or replacement of the ATAES132A device with a malicious counterpart (AV:P per CVSS vector). The attacker must also hold low-level privileges on the system (PR:L per CVSS). The Zephyr build must include the ATAES132A driver (CONFIG_CRYPTO_ATAES132A enabled), which is not a default driver in general-purpose Zephyr builds and requires the target hardware to incorporate the ATAES132A chip. Attack complexity is High (AC:H), indicating that exploit reliability is low and likely requires precise knowledge of stack layout and return address offsets. Exploitation is not remotely triggerable and is bounded to hardware designs that incorporate the ATAES132A co-processor.

The CVSS 3.1 base score of 3.8 (Low) is largely driven by AV:P (physical access to the device bus required) and AC:H (high attack complexity), which dramatically suppress the real-world attack surface. EPSS at 0.01% in the 3rd percentile corroborates that observed exploitation probability is negligible. The vulnerability is absent from CISA KEV and no public proof-of-concept has been identified, meaning no automation of exploitation has been demonstrated. Despite these constrained signals, the underlying primitive - a classic stack buffer overflow in kernel space with kernel memory corruption and potential execution hijack - is technically severe if exploitation preconditions are met. The vendor's conservative impact scoring (C:L/I:L/A:L) likely reflects the difficulty of reliable exploitation rather than the theoretical ceiling of a successful attack. This is a case where real-world risk closely tracks CVSS: the physical and complexity barriers make this a low-priority item for most deployments unless the threat model explicitly includes hardware supply-chain or physical-access adversaries. SSVC assessment: exploitation status is 'none,' the vulnerability is not automatable given AV:P, and technical impact is partial-to-total depending on exploit reliability - overall decision: track and patch on normal update cadence.

An attacker with physical access to the I2C or SPI bus on a Zephyr-based IoT device - either via a hardware bus tap or by substituting a malicious ATAES132A-compatible device - sends a crafted response packet with an oversized length field to the Zephyr crypto driver. The driver copies the oversized payload into the 52-byte stack buffer without bounds checking, overflowing into adjacent kernel stack frames and overwriting saved return addresses. With precise control over the overflow payload and knowledge of the stack layout, the attacker redirects kernel execution to attacker-controlled code. No public proof-of-concept has been identified at time of analysis.