Automatic speech recognition (ASR) is popular in our daily lives (e.g., via voice assistants or voice input). Once its security attributes are destroyed, it poses as a severe threat to a user’s life and ‘property safety’. Prior research has demonstrated that ASR systems are vulnerable to backdoor attacks. A model embedded with a backdoor behaves normally on clean samples yet misclassifies malicious samples that contain triggers. Existing backdoor attacks have mostly been conducted in the image domain. However, they can not be applied in the audio domain because of poor transferability. This paper proposes a dynamic backdoor attack method against ASR models, named DriNet. Significantly, we designed a dynamic trigger generation network to craft a variety of audio triggers. It is trained jointly with the discriminative model incorporated with an attack success rate on poisoned samples and accuracy on clean samples. We demonstrate that DriNet achieves an attack success rate of 86.4% when infecting only 0.5% of the training set without reducing its accuracy. DriNet can still achieve comparable attack performance to backdoor attacks using static triggers, further enjoying richer attack patterns. We further evaluated DriNet’s resistance to a current state-of-the-art defense mechanism. The anomaly index of DriNet is more than 37.4% smaller than that of BadNets method. The triggers generated by DriNet are hard reverse, keeping DriNet from the detectors.

Sarcasm detection remains a challenge for numerous Natural Language Processing (NLP) tasks, such as sentiment classification or stance prediction. Existing sarcasm detection studies attempt to capture the subtle semantic incongruity patterns by using contextual information and graph information through Graph Convolutional Networks (GCN). However, direct application of dependence may inevitably introduce noisy information and inferiorly in modeling long-distance or disconnected words in the dependency tree. To better learn the sentiment inconsistencies between terms, we propose an Affection Enhanced Relational Graph Attention network (ARGAT) by jointly considering the affective information and the dependency information. Specifically, we use Relational Graph Attention Networks (RGAT) to integrate relation information guided by a trainable matrix of relation types and synchronously use GCNs to integrate affection information explicitly donated by affective adjacency matrixes. The employment of RGAT contributes to information interaction of structural relevant word pairs with a long distance. With the enhancement of affective information, the proposed model can capture complex forms of sarcastic expressions. Experimental results on six benchmark datasets show that our proposed approach outperforms state-of-the-art sarcasm detection methods. The best-improved results of accuracy and F1 are 4.19% and 4.33%, respectively.