Estimating Vulnerability of All Model Parameters in DNN with a Small Number of Fault Injections
Estimating Vulnerability of All Model Parameters in DNN with a Small Number of Fault Injections
复制标题
DOI：
10.23919/date54114.2022.9774569
复制
发表时间：
2022-03
期刊：
影响因子：
--
通讯作者：
Yangchao Zhang;Hiroaki Itsuji;T. Uezono;Tadanobu Toba;Masanori Hashimoto
中科院分区：
文献类型：
--
作者：
Yangchao Zhang;Hiroaki Itsuji;T. Uezono;Tadanobu Toba;Masanori Hashimoto
随着深度神经网络越来越多地应用于自动驾驶等安全关键应用，深度神经网络（dnn）对硬件错误的可靠性至关重要。内存中的瞬态错误，如辐射引起的软错误，可能会通过推理计算传播，导致意外输出，这可能会引发灾难性的系统故障。作为解决这一问题的第一步，本文提出构建一个带有少量故障注入的漏洞模型（VM）来识别深度神经网络中的漏洞模型参数。我们大大减少了故障注入的位位数，并开发了一个流程来增量收集训练数据，即故障注入结果，以提高VM的精度。实验结果表明，与传统的基于故障注入的漏洞估计相比，虚拟机仅用1/3490的计算量即可估计出所有DNN模型参数的漏洞。
The reliability of deep neural networks (DNNs) against hardware errors is essential as DNNs are increasingly employed in safety-critical applications such as automatic driving. Transient errors in memory, such as radiation-induced soft error, may propagate through the inference computation, resulting in unexpected output, which can adversely trigger catastrophic system failures. As a first step to tackle this problem, this paper proposes constructing a vulnerability model (VM) with a small number of fault injections to identify vulnerable model parameters in DNN. We reduce the number of bit locations for fault injection significantly and develop a flow to incrementally collect the training data, i.e., the fault injection results, for VM accuracy improvement. Experimental results show that VM can estimate vulnerabilities of all DNN model parameters only with 1/3490 computations compared with traditional fault injection-based vulnerability estimation.