DEMIX: Domain-Enforced Memory Isolation for Embedded System.

Area of Science:

• Computer Science
• Embedded Systems Engineering
• Cybersecurity

Background:

• Memory isolation is crucial for securing lightweight embedded systems.
• The Memory Protection Unit (MPU) is common but incurs overheads during domain switching.
• Increasing IoT complexity necessitates more granular memory protection without performance penalties.

Purpose of the Study:

• To propose DEMIX, a novel technique for efficient memory isolation in multi-domain embedded systems.
• To eliminate the overhead associated with traditional domain switching methods.
• To enable fine-grained memory access control and extend domain capabilities.

Main Methods:

• DEMIX utilizes Domain-Enforced Memory Isolation and instruction-level domain isolation.
• It validates the processor's domain state and executed instructions for granular access control.
• Implementation and performance evaluation on a lightweight RISC-V processor (Ibex Core).

Main Results:

• DEMIX successfully supports efficient memory isolation for multiple domains.
• The technique eliminates domain switching overheads.
• Implementation on Ibex Core resulted in only an 8% hardware overhead for eight user domains.

Conclusions:

• DEMIX provides a secure and efficient solution for memory isolation in embedded systems.
• It overcomes the limitations of traditional MPU-based approaches by removing domain switching overhead.
• The proposed method enables enhanced security and scalability for IoT environments.