[Paper Review] Evaluating the Efficiency of Software-only Techniques to Detect SEU and SET in Microprocessors

José Rodrigo Azambuja, Fernando Sousa+2 more|arXiv (Cornell University)|Sep 28, 2023

Radiation Effects in Electronics12 citations

TL;DR

The paper evaluates software-only rules to harden a MIPS-based microprocessor against SEU and SET, finding many control-flow techniques inefficient for detecting faults.

ABSTRACT

This paper presents a detailed evaluation of the efficiency of software-only techniques to mitigate SEU and SET in microprocessors. A set of well-known rules is presented and implemented automatically to transform an unprotected program into a hardened one. SEU and SET are injected in all sensitive areas of a MIPS-based microprocessor architecture. The efficiency of each rule and a combination of them are tested. Experimental results show the inefficiency of the control-flow techniques in detecting the majority of SEU and SET faults. Three effects of the non-detected faults are explained. The conclusions can lead designers in developing more efficient techniques to detect these types of faults.

Motivation & Objective

• Motivate the need for software-based mitigation techniques against SEU and SET in microprocessors.
• Present a set of well-known rules that can automatically transform an unprotected program into a hardened version.
• Evaluate the efficiency of each rule and combinations of rules through fault injection in a MIPS-based architecture.
• Explain the implications of non-detected faults for hardware designers and motivate improved techniques.

Proposed method

• Automatically implement a set of well-known software-based rules to transform unprotected code into hardened code.
• Inject SEU and SET faults into all sensitive areas of a MIPS-based microprocessor to assess detection efficacy.
• Evaluate the efficiency of each rule individually and in combinations.
• Analyze the reasons for non-detection and the resulting fault effects.

Experimental results

Research questions

• RQ1 How effective are individual software-only rules at detecting SEU and SET faults in a MIPS-based processor?
• RQ2 Do combinations of software-based rules improve detection coverage for SEU and SET faults?
• RQ3 What are the main failure modes when software-only techniques fail to detect faults?

Key findings

• Control-flow based techniques detect only a minority of SEU and SET faults, showing inefficiency.
• Multiple non-detected faults exhibit three distinct effects (as described in the results).
• The results suggest designers should pursue more effective or complementary techniques beyond control-flow rules.