Search for: All records

Speleothem δ¹⁸O records from central southern China have long been regarded as a key benchmark for Asian summer monsoon intensity. However, the similar δ¹⁸O minima observed among precession minima and their link to seasonal precipitation mixing remains unclear. Here, we present a 400,000-y record of summer precipitation δ¹⁸O from loess microcodium, which captures distinct precession cycles similar to those seen in speleothem δ¹⁸O records, particularly during glacial periods. Notably, our microcodium δ¹⁸O record reveals very low-δ¹⁸O values during precession minima at peak interglacials, a feature absent in speleothem δ¹⁸O records from central southern China. This discrepancy suggests that the mixed summer and nonsummer climatic signals substantially influence the speleothem δ¹⁸O records from central southern China. Proxy-model comparisons indicate that the lack of very low-δ¹⁸O values in speleothem δ¹⁸O records is due to an attenuated summer signal contribution, resulting from a lower summer-to-annual precipitation ratio in southern China at strong monsoon intervals. Our findings offer a potential explanation for the long-standing puzzle of the absence of 100- and 41-kyr cycles in speleothem δ¹⁸O records and underscore the critical role of seasonality in interpreting paleoclimatic proxies in central southern China. These insights also have broader implications for interpreting speleothem δ¹⁸O records globally, advocating for a more multiseason interpretive framework. 

As the real-world applications (image segmentation, speech recognition, machine translation, etc.) are increasingly adopting Deep Neural Networks (DNNs), DNN's vulnerabilities in a malicious environment have become an increasingly important research topic in adversarial machine learning. Adversarial machine learning (AML) focuses on exploring vulnerabilities and defensive techniques for machine learning models. Recent work has shown that most adversarial audio generation methods fail to consider audios' temporal dependency (TD) (i.e., adversarial audios exhibit weaker TD than benign audios). As a result, the adversarial audios are easily detectable by examining their TD. Therefore, one area of interest in the audio AML community is to develop a novel attack that evades a TD-based detection model. In this contribution, we revisit the LSTM model for audio transcription and propose a new audio attack algorithm that evades the TD-based detection by explicitly controlling the TD in generated adversarial audios. The experimental results show that the detectability of our adversarial audio is significantly reduced compared to the state-of-the-art audio attack algorithms. Furthermore, experiments also show that our adversarial audios remain nearly indistinguishable from benign audios with only negligible perturbation magnitude. 

The robustness and vulnerability of Deep Neural Networks (DNN) are quickly becoming a critical area of interest since these models are in widespread use across real-world applications (i.e., image and audio analysis, recommendation system, natural language analysis, etc.). A DNN's vulnerability is exploited by an adversary to generate data to attack the model; however, the majority of adversarial data generators have focused on image domains with far fewer work on audio domains. More recently, audio analysis models were shown to be vulnerable to adversarial audio examples (e.g., speech command classification, automatic speech recognition, etc.). Thus, one urgent open problem is to detect adversarial audio reliably. In this contribution, we incorporate a separate and yet related DNN technique to detect adversarial audio, namely model quantization. Then we propose an algorithm to detect adversarial audio by using a DNN's quantization error. Specifically, we demonstrate that adversarial audio typically exhibits a larger activation quantization error than benign audio. The quantization error is measured using character error rates. We use the difference in errors to discriminate adversarial audio. Experiments with three the-state-of-the-art audio attack algorithms against the DeepSpeech model show our detection algorithm achieved high accuracy on the Mozilla dataset.