In this paper, we show that stress-tests can be potentially used as power-noise viruses in denial-of-service (DoS) attacks by causing voltage emergencies that may lead to data corruptions and system crashes in multi-core processors. This attack targets processors whose operating voltage has been reduced in-the-field for improving energy efficiency. To protect such undervolted processors from this type of attacks, we present a run-time system for detecting and mitigating power-noise viruses. We present voltage noise data from power-noise viruses and benchmarks collected from an Arm multi-core processor, and we observe that the frequency of voltage emergencies dramatically increases during the execution of power-noise attacks. Based on this observation, we propose a regression model that allows for a run-time estimation of the severity of voltage emergencies by monitoring the frequency of voltage emergencies and the operating frequency of the processor. For mitigating the problem, during the execution of critical tasks requiring protection, our system periodically evaluates the severity of voltage emergencies and adapts the operating frequency of the processor in order to reduce the severity of the attack according to a predefined constraint. We demonstrate the efficiency of the proposed run-time protection system on an actual Arm multi-core processor using two power-noise viruses, and we explore trade-offs between protection latency, CPU utilization and power cost. The proposed software achieves with a very low CPU utilization overhead of less than 0.11% to detect and mitigate power-noise DoS attacks with a latency of 100 $\mu {\mathrm{ s}}$ .

中文翻译：

---

多核处理器的运行时保护免受电源噪声拒绝服务攻击

在本文中，我们表明压力测试可能会在拒绝服务 (DoS) 攻击中用作电源噪声病毒，因为它会引起电压紧急情况，从而可能导致多核处理器中的数据损坏和系统崩溃。这种攻击的目标是在现场降低工作电压以提高能效的处理器。为了保护这种低电压处理器免受此类攻击，我们提出了一个运行时系统来检测和减轻电源噪声病毒。我们提供了来自电源噪声病毒的电压噪声数据和从 Arm 多核处理器收集的基准测试，我们观察到在电源噪声攻击执行期间电压紧急情况的频率急剧增加。基于这一观察，我们提出了一个回归模型，该模型允许通过监控电压紧急情况的频率和处理器的工作频率来对电压紧急情况的严重程度进行运行时估计。为了缓解问题，在执行需要保护的关键任务期间，我们的系统会定期评估电压紧急情况的严重程度并调整处理器的工作频率，以便根据预定义的约束降低攻击的严重程度。我们使用两种电源噪声病毒在实际的 Arm 多核处理器上演示了所提议的运行时保护系统的效率，并探索了保护延迟、CPU 利用率和电源成本之间的权衡。建议的软件以低于 0 的非常低的 CPU 利用率开销实现。