Multi-Channel Multi-Scale Convolution Attention Variational Autoencoder (MCA-VAE): An Interpretable Anomaly Detection Algorithm Based on Variational Autoencoder.

With the rapid development of industry, the risks factories face are increasing. Therefore, the anomaly detection algorithms deployed in factories need to have high accuracy, and they need to be able to promptly discover and locate the specific equipment causing the anomaly to restore the regular operation of the abnormal equipment. However, the neural network models currently deployed in factories cannot effectively capture both temporal features within dimensions and relationship features between dimensions; some algorithms that consider both types of features lack interpretability. Therefore, we propose a high-precision, interpretable anomaly detection algorithm based on variational autoencoder (VAE). We use a multi-scale local weight-sharing convolutional neural network structure to fully extract the temporal features within each dimension of the multi-dimensional time series. Then, we model the features from various aspects through multiple attention heads, extracting the relationship features between dimensions. We map the attention output results to the latent space distribution of the VAE and propose an optimization method to improve the reconstruction performance of the VAE, detecting anomalies through reconstruction errors. Regarding anomaly interpretability, we utilize the VAE probability distribution characteristics, decompose the obtained joint probability density into conditional probabilities on each dimension, and calculate the anomaly score, which provides helpful value for technicians. Experimental results show that our algorithm performed best in terms of F1 score and AUC value. The AUC value for anomaly detection is 0.982, and the F1 score is 0.905, which is 4% higher than the best-performing baseline algorithm, Transformer with a Discriminator for Anomaly Detection (TDAD). It also provides accurate anomaly interpretation capability.

Malware Identification Method in Industrial Control Systems Based on Opcode2vec and CVAE-GAN.

Industrial Control Systems (ICSs) have faced a significant increase in malware threats since their integration with the Internet. However, existing machine learning-based malware identification methods are not specifically optimized for ICS environments, resulting in suboptimal identification performance. In this work, we propose an innovative method explicitly tailored for ICSs to enhance the performance of malware classifiers within these systems. Our method integrates the opcode2vec method based on preprocessed features with a conditional variational autoencoder-generative adversarial network, enabling classifiers based on Convolutional Neural Networks to identify malware more effectively and with some degree of increased stability and robustness. Extensive experiments validate the efficacy of our method, demonstrating the improved performance of malware classifiers in ICSs. Our method achieved an accuracy of 97.30%, precision of 92.34%, recall of 97.44%, and F1-score of 94.82%, which are the highest reported values in the experiment.

A Lightweight Unsupervised Intrusion Detection Model Based on Variational Auto-Encoder.

With the gradual integration of internet technology and the industrial control field, industrial control systems (ICSs) have begun to access public networks on a large scale. Attackers use these public network interfaces to launch frequent invasions of industrial control systems, thus resulting in equipment failure and downtime, production data leakage, and other serious harm. To ensure security, ICSs urgently need a mature intrusion detection mechanism. Most of the existing research on intrusion detection in ICSs focuses on improving the accuracy of intrusion detection, thereby ignoring the problem of limited equipment resources in industrial control environments, which makes it difficult to apply excellent intrusion detection algorithms in practice. In this study, we first use the spectral residual (SR) algorithm to process the data; we then propose the improved lightweight variational autoencoder (LVA) with autoregression to reconstruct the data, and we finally perform anomaly determination based on the permutation entropy (PE) algorithm. We construct a lightweight unsupervised intrusion detection model named LVA-SP. The model as a whole adopts a lightweight design with a simpler network structure and fewer parameters, which achieves a balance between the detection accuracy and the system resource overhead. Experimental results on the ICSs dataset show that our proposed LVA-SP model achieved an F1-score of 84.81% and has advantages in terms of time and memory overhead.

An Access Control System Based on Blockchain with Zero-Knowledge Rollups in High-Traffic IoT Environments.

The access control (AC) system in an IoT (Internet of Things) context ensures that only authorized entities have access to specific devices and that the authorization procedure is based on pre-established rules. Recently, blockchain-based AC systems have gained attention within research as a potential solution to the single point of failure issue that centralized architectures may bring. Moreover, zero-knowledge proof (ZKP) technology is included in blockchain-based AC systems to address the issue of sensitive data leaking. However, current solutions have two problems: (1) systems built by these works are not adaptive to high-traffic IoT environments because of low transactions per second (TPS) and high latency; (2) these works cannot fully guarantee that all user behaviors are honest. In this work, we propose a blockchain-based AC system with zero-knowledge rollups to address the aforementioned issues. Our proposed system implements zero-knowledge rollups (ZK-rollups) of access control, where different AC authorization requests can be grouped into the same batch to generate a uniform ZKP, which is designed specifically to guarantee that participants can be trusted. In low-traffic environments, sufficient experiments show that the proposed system has the least AC authorization time cost compared to existing works. In high-traffic environments, we further prove that based on the ZK-rollups optimization, the proposed system can reduce the authorization time overhead by 86%. Furthermore, the security analysis is presented to show the system's ability to prevent malicious behaviors.

High-Speed Extraction of Regions of Interest in Optical Camera Communication Enabled by Grid Virtual Division.

Optical camera communication (OCC), enabled by light-emitting diodes (LEDs) and embedded cameras on smartphones, has drawn considerable attention thanks to the pervasive adoption of LED lighting and mobile devices. However, most existing studies do not consider the performance bottleneck of Region of Interest (RoI) extraction during decoding, making it challenging to improve communication capacity further. To this end, we propose a fast grid virtual division scheme based on pixel grayscale values, which extracts RoI quickly without sacrificing computational complexity, thereby reducing the decoding delay and improving the communication capacity of OCC. Essentially, the proposed scheme uses a grid division strategy to divide the received image into blocks and randomly sample several pixels within different blocks to quickly locate the RoI with high grayscale values in the original image. By implementing the lightweight RoI extraction algorithm, we experimentally verify its effectiveness in reducing decoding latency, demonstrating its superior performance in terms of communication capacity. The experimental results clearly show that the decoding delay of the proposed scheme is 70% lower than that provided by the Gaussian blur scheme for the iPhone receiver at a transmission frequency of 5 kHz.

DeepEdge: A Novel Appliance Identification Edge Platform for Data Gathering, Capturing and Labeling.

With the development of the Internet of Things for smart grid, the requirement for appliance monitoring has become an important topic. The first and most important step in appliance monitoring is to identify the type of appliance. Most of the existing appliance identification platforms are cloud based, thus they consume large computing resources and memory. Therefore, it is necessary to explore an edge identification platform with a low cost. In this work, a novel appliance identification edge platform for data gathering, capturing and labeling is proposed. Experiments show that this platform can achieve an average appliance identification accuracy of 98.5% and improve the accuracy of non-intrusive load disaggregation algorithms.

Semi-Supervised Instance-Segmentation Model for Feature Transfer Based on Category Attention.

In the task of image instance segmentation, semi-supervised instance segmentation algorithms have received constant research attention over recent years. Among these algorithms, algorithms based on transfer learning are better than algorithms based on pseudo-label generation in terms of segmentation performance, but they can not make full use of the relevant characteristics of source tasks. To improve the accuracy of these algorithms, this work proposes a semi-supervised instance segmentation model AFT-Mask (attention-based feature transfer Mask R-CNN) based on category attention. The AFT-Mask model takes the result of object-classification prediction as "attention" to improve the performance of the feature-transfer module. In detail, we designed a migration-optimization module for connecting feature migration and classification prediction to enhance segmentation-prediction accuracy. To verify the validity of the AFT-Mask model, experiments were conducted on two types of datasets. Experimental results show that the AFT-Mask model can achieve effective knowledge transfer and improve the performance of the benchmark model on semi-supervised instance segmentation.

High-Accuracy Height-Independent 3D VLP Based on Received Signal Strength Ratio.

Visible light positioning (VLP) has attracted intensive attention from both academic and industrial communities thanks to its high accuracy, immunity to electromagnetic interference, and low deployment cost. In general, the receiver in a VLP system determines its own position by exploring the received signal strength (RSS) from the transmitter according to a pre-built RSS attenuation model. In such model-based methods, the LED's emission power and the receiver's height are usually required known and constant parameters to obtain reasonable positioning accuracy. However, the LED's emission power is normally time-varying due to the fact that the LED's optical output power is prone to changing with the LED's temperature, and the receiver's height is random in a realistic application scenario. To this end, we propose a height-independent three-dimensional (3D) VLP scheme based on the RSS ratio (RSSR), rather than only using RSS. Unlike existing RSS-based VLP methods, our method is able to independently find the horizontal coordinate, i.e., two-dimensional (2D) position, without a priori height information of the receiver, and also avoids the negative effect caused by fluctuation of the LED's emission power. Moreover, we can further infer the height of the receiver to achieve three-dimensional (3D) positioning by iterating the 2D results back into positioning equations. To quickly verify the proposed scheme, we conduct theoretical analysis with mathematical proof and experimental results with real data, which confirm that the proposed scheme can achieve high position accuracy without known information of the receiver's height and LED's emission power. We also implement a VLP prototype with five LED transmitters, and experimental results show that the proposed scheme can achieve very low average errors of 2.73 cm in 2D and 7.20 cm in 3D.

Intelligent Reflecting Surface-Assisted Physical Layer Key Generation with Deep Learning in MIMO Systems.

Physical layer secret key generation (PLKG) is a promising technology for establishing effective secret keys. Current works for PLKG mostly study key generation schemes in ideal communication environments with little or even no signal interference. In terms of this issue, exploiting the reconfigurable intelligent reflecting surface (IRS) to assist PLKG has caused an increasing interest. Most IRS-assisted PLKG schemes focus on the single-input-single-output (SISO), which is limited in future communications with multi-input-multi-output (MIMO). However, MIMO could bring a serious overhead of channel reciprocity extraction. To fill the gap, this paper proposes a novel low-overhead IRS-assisted PLKG scheme with deep learning in the MIMO communications environments. We first combine the direct channel and the reflecting channel established by the IRS to construct the channel response function, and we propose a theoretically optimal interaction matrix to approach the optimal achievable rate. Then we design a channel reciprocity-learning neural network with an IRS introduced (IRS-CRNet), which is exploited to extract the channel reciprocity in time division duplexing (TDD) systems. Moreover, a PLKG scheme based on the IRS-CRNet is proposed. Final simulation results verify the performance of the PLKG scheme based on the IRS-CRNet in terms of key generation rate, key error rate and randomness.

Enhancing the performance of optical camera communication via accumulative sampling.

Deemed as a practical approach to realize Visible Light Communication on commercial-off-the-shelf devices, the Optical Camera Communication (OCC) is attracting increasing attention, thanks to its readiness to be built purely upon ubiquitous LED illuminating infrastructure and handy smartphones. However, limited by the low sampling ability of the built-in camera on a smartphone, the performance of existing OCC systems is still far away from the requirements of practical applications. To this end, we further investigate the reception ability of the smartphone's camera and propose an accumulative sampling scheme to improve the performance of the OCC system. Essentially, the proposed scheme can use all the grayscale information of the pixels projected by the LED transmitter, whereas the conventional ones normally use single row (or column) pixels for demodulating. By implementing the lightweight demodulation algorithm with accumulative sampling, we experimentally verify its effectiveness for supporting higher transmission frequency hence better performance in terms of data rate. Extensive evaluations have shown the BERs of the proposed method are over 87% and 96% lower than that provided by the baselines at a maximum transmission frequency of 5 kHz for the Samsung S8 and iPhone 8 Plus receivers, respectively.

Improved optical camera communication systems using a freeform lens.

Optical camera communication (OCC) systems, which utilize image sensors embedded in commercial-off-the-shelf devices to detect time and spatial variations in light intensity for enabling data communications, have stirred up researchers' interest. Compared to a direct OCC system whose maximum data rate is strongly determined by the LED source size, a reflected OCC system can break that limitation since the camera captures the light rays reflecting off an observation plane (e.g., a wall) instead of those light rays directly emanated from the light source. However, the low signal-to-noise ratio caused by the non-uniform irradiance distribution produced by LED luminaire on the observation plane in current reflected OCC systems cannot be avoided, hence low complexity and accurate demodulation are hard to achieve. In this paper, we present a FreeOCC system, which employs a dedicatedly tailored freeform lens to precisely control the propagation of modulated light. A desired uniform rectangular illumination is produced on the observation plane by the freeform lens, yielding a uniform grayscale distribution within the received frame captured by the camera in the proposed FreeOCC system. Then, the received signal can be easily demodulated with high accuracy by a simple thresholding scheme. A prototype of the FreeOCC system demonstrates the high performance of the proposed system, and two pulse amplitude modulation schemes (4-order and 8-order) are performed. By using the freeform lens, the packet reception rate is increased by 35% and 32%, respectively; the bit error rate is decreased by 72% and 59%, respectively, at a transmission frequency of 5 kHz. The results clearly show that the FreeOCC system outperforms the common reflected OCC system.

HAGP: A Heuristic Algorithm Based on Greedy Policy for Task Offloading with Reliability of MDs in MEC of the Industrial Internet.

In the Industrial Internet, computing- and power-limited mobile devices (MDs) in the production process can hardly support the computation-intensive or time-sensitive applications. As a new computing paradigm, mobile edge computing (MEC) can almost meet the requirements of latency and calculation by handling tasks approximately close to MDs. However, the limited battery capacity of MDs causes unreliable task offloading in MEC, which will increase the system overhead and reduce the economic efficiency of manufacturing in actual production. To make the offloading scheme adaptive to that uncertain mobile environment, this paper considers the reliability of MDs, which is defined as residual energy after completing a computation task. In more detail, we first investigate the task offloading in MEC and also consider reliability as an important criterion. To optimize the system overhead caused by task offloading, we then construct the mathematical models for two different computing modes, namely, local computing and remote computing, and formulate task offloading as a mixed integer non-linear programming (MINLP) problem. To effectively solve the optimization problem, we further propose a heuristic algorithm based on greedy policy (HAGP). The algorithm achieves the optimal CPU cycle frequency for local computing and the optimal transmission power for remote computing by alternating optimization (AP) methods. It then makes the optimal offloading decision for each MD with a minimal system overhead in both of these two modes by the greedy policy under the limited wireless channels constraint. Finally, multiple experiments are simulated to verify the advantages of HAGP, and the results strongly confirm that the considered task offloading reliability of MDs can reduce the system overhead and further save energy consumption to prolong the life of the battery and support more computation tasks.

Optical-spatial-summing-based NOMA with fine-grained power allocation for VLC-enabled IoT applications.

In this Letter, we propose and demonstrate a practical optical-spatial-summing-based non-orthogonal multiple access (OSS-NOMA) technique for visible light communication (VLC) systems. This technique is innovative in adopting OSS in that the transmitter of OSS-NOMA VLC can be built upon commercial illuminating light emitting diodes (LEDs), free of LEDs' harmful nonlinearity. Unlike conventional NOMA VLC using analog components such as digital-to-analog converters and bias-T in the transmitter side, OSS-NOMA exploits only digital control signals to drive a LED array in forming optical power superposition for NOMA signals. We demonstrate that by simply switching different amounts of LED chips on or off, the proposed OSS-NOMA transmitter can deliver a fine-grained power allocation ratio ranging from 0.01 to one for two users. The implemented OSS-NOMA VLC prototype leveraging commercial components can achieve low bit error rates of ≤3.1×10-3 for two users at a data rate of 800 kbps, confirming the promising potential of this novel OSS-NOMA VLC for Internet of Things (IoT) applications.

A Safe Charging Algorithm Based on Multiple Mobile Chargers.

A safe charging algorithm in wireless rechargeable sensor network ensures the charging efficiency and the electromagnetic radiation below the threshold. Compared with the current charging algorithms, the safe charging algorithm is more complicated due to the radiation constraint and the mobility of the chargers. A safe charging algorithm based on multiple mobile chargers is proposed in this paper to charge the sensor nodes with mobile chargers, in order to ensure the premise of radiation safety, multiple mobile chargers can effectively complete the network charging task. Firstly, this algorithm narrows the possible location of the sensor nodes by utilizing the charging time and antenna waveform. Secondly, the performance of non-partition charging algorithm which algorithm allow chargers to charge different sensors sets in a different cycle is evaluated against the one of partition charging which does not allow for charging different ones. The moving distance of the charger node will be reduced by 18%. It not only improves the safety level which is inversely proportional to electromagnetic radiation but also expands the application scope of the wireless sensor nodes.

Low complexity OFDM VLC system enabled by spatial summing modulation.

Commercial-off-the-shelf (COTS) devices enabled visible light communication (VLC) for Internet of things (IoT) applications has attracted extensive attentions from both academic and industrial communities, thanks to the pervasive deployments of light emitting diode (LED) lighting infrastructure. However, due to the limitation of frequency response and non-linearity of the commercial illuminating LED light consisting of multiple LED chips, the achievable data rate is far less than that provided by the experimental VLC system with a single LED with specialized devices, e.g., lens. To this end, we propose a power-of-2 arrangement scheme for LED chips to generate spatial summing modulation with low control complexity, and demonstrate its availability in an orthogonal frequency division multiplexing (OFDM) VLC system purely built upon COTS devices. It significantly differs from a conventional OFDM VLC system relying on digital-to-analog converter (DAC) and analog signal chain, which is complex and confined by LED's non-linearity, thanks to we design a novel digital-to-light converter (DLC) which can output 256 light intensities linearly and be directly controlled by the discrete digital signals generated by the OFDM modulator. An experimental demonstration with employing the QAM-OFDM modulation scheme successfully confirms the effectiveness of the proposed spatial summing VLC system, which can achieve low BERs of below the forward error correct (FEC) threshold of 3.8×10-3 for both QAM8 and QAM16 running transmission frequency of 300 kHz under a communication distance of 0.8 m. It demonstrates the promising potential for delivering a data rate at hundred kbps level with this novel spatial summing based OFDM VLC system, which is sufficient for many IoT applications.

Automatic detection on intracranial aneurysm from digital subtraction angiography with cascade convolutional neural networks.

BACKGROUND: An intracranial aneurysm is a cerebrovascular disorder that can result in various diseases. Clinically, diagnosis of an intracranial aneurysm utilizes digital subtraction angiography (DSA) modality as gold standard. The existing automatic computer-aided diagnosis (CAD) research studies with DSA modality were based on classical digital image processing (DIP) methods. However, the classical feature extraction methods were badly hampered by complex vascular distribution, and the sliding window methods were time-consuming during searching and feature extraction. Therefore, developing an accurate and efficient CAD method to detect intracranial aneurysms on DSA images is a meaningful task. METHODS: In this study, we proposed a two-stage convolutional neural network (CNN) architecture to automatically detect intracranial aneurysms on 2D-DSA images. In region localization stage (RLS), our detection system can locate a specific region to reduce the interference of the other regions. Then, in aneurysm detection stage (ADS), the detector could combine the information of frontal and lateral angiographic view to identify intracranial aneurysms, with a false-positive suppression algorithm. RESULTS: Our study was experimented on posterior communicating artery (PCoA) region of internal carotid artery (ICA). The data set contained 241 subjects for model training, and 40 prospectively collected subjects for testing. Compared with the classical DIP method which had an accuracy of 62.5% and an area under curve (AUC) of 0.69, the proposed architecture could achieve accuracy of 93.5% and the AUC of 0.942. In addition, the detection time cost of our method was about 0.569 s, which was one hundred times faster than the classical DIP method of 62.546 s. CONCLUSION: The results illustrated that our proposed two-stage CNN-based architecture was more accurate and faster compared with the existing research studies of classical DIP methods. Overall, our study is a demonstration that it is feasible to assist physicians to detect intracranial aneurysm on DSA images using CNN.

VN-NDP: A Neighbor Discovery Protocol Based on Virtual Nodes in Mobile WSNs.

As an indispensable part of Internet of Things (IoT), wireless sensor networks (WSNs) are more and more widely used with the rapid development of IoT. The neighbor discovery protocols are the premise of communication between nodes and networking in energy-limited self-organizing wireless networks, and play an important role in WSNs. Because the node energy is limited, neighbor discovery must operate in an energy-efficient manner, that is, under the condition of a given energy budget, the neighbor discovery performance should be as good as possible, such that the discovery latency would be as small as possible and the discovered neighbor percentage as large as possible. The indirect neighbor discovery mainly uses the information of the neighbors that have been found by a pairwise discovery method to more efficiently make a re-planning of the discovery wake-up schedules of the original pairwise neighbor discovery, thereby improving the discovery energy efficiency. The current indirect neighbor discovery methods are mainly divided into two categories: one involves removing the inefficient active slots in the original discovery wake-up schedules, and the other involves adding some efficient active slots. However, the two categories of methods have their own limitations. The former does not consider that this removal operation destroys the integrity of the original discovery wake-up schedules and hence the possibility of discovering new neighbors is reduced, which adversely affects the discovered neighbor percentage. For the latter category, there are still inefficient active slots that were not removed in the re-planned wake-up schedules. The motivation of this paper is to combine the advantages of these two types of indirect neighbor discovery methods, that is, to combine the addition of efficient active slots and the removal of inefficient active slots. To achieve this goal, this paper proposes, for the first time, the concept of virtual nodes in neighbor discovery to maximize the integrity of the original wake-up schedules and achieve the goals of adding efficient active slots and removing inefficient active slots. Specifically, a virtual node is a collaborative group that is formed by nodes within a small range. The nodes in a collaborative group share responsibility for the activating task of one member node, and the combination of these nodes' wake-up schedules forms the full wake-up schedule of a node that only uses a pairwise method. In addition, this paper proposes a set of efficient group management mechanisms, and the key steps affecting energy efficiency are analyzed theoretically to obtain the energy-optimal parameters. The extended simulation experiments in multiple scenarios show that, compared with other methods, our neighbor discovery protocol based on virtual nodes (VN-NDP) has a significant improvement in average discovery delay and discovered neighbor percentage performance at a given energy budget. Compared with the typical indirect neighbor discovery algorithm EQS, a neighbor discovery with extended quorum system, our proposed VN-NDP method reduces the average discovery delay by up to 10 . 03 % and increases the discovered neighbor percentage by up to 18 . 35 % .

Secure and private NOMA VLC using OFDM with two-level chaotic encryption.

In this paper, we propose and demonstrate a secure and private non-orthogonal multiple access (NOMA) based visible light communication (VLC) system. Orthogonal frequency division multiplexing (OFDM) modulation is applied in the system and a two-level chaotic encryption scheme is further implemented, which can guarantee both the security of legitimate users against eavesdroppers and the privacy among all the legitimate users. An experimental demonstration with two legitimate users and one eavesdropper successfully verifies the feasibility of the proposed secure and private NOMA VLC system. To the best of our knowledge, it is the first time that simultaneous security and privacy improvement is considered for NOMA VLC systems.

A Fast Neighbor Discovery Algorithm in WSNs.

With the quick development of Internet of Things (IoT), one of its important supporting technologies, i.e., wireless sensor networks (WSNs), gets much more attention. Neighbor discovery is an indispensable procedure in WSNs. The existing deterministic neighbor discovery algorithms in WSNs ensure that successful discovery can be obtained within a given period of time, but the average discovery delay is long. It is difficult to meet the need for rapid discovery in mobile low duty cycle environments. In addition, with the rapid development of IoT, the node densities of many WSNs greatly increase. In such scenarios, existing neighbor discovery methods fail to satisfy the requirement in terms of discovery latency under the condition of the same energy consumption. This paper proposes a group-based fast neighbor discovery algorithm (GBFA) to address the issues. By carrying neighbor information in beacon packet, the node knows in advance some potential neighbors. It selects more energy efficient potential neighbors and proactively makes nodes wake up to verify whether these potential neighbors are true neighbors, thereby speeding up neighbor discovery, improving energy utilization efficiency and decreasing network communication load. The evaluation results indicate that, compared with other methods, GBFA decreases the average discovery latency up to 10 . 58 % at the same energy budget.

Community Composition of Photosynthetic Picoeukaryotes in a Subtropical Coastal Ecosystem, with Particular Emphasis on Micromonas.

Photosynthetic picoeukaryotes (PPEs) are important constituents in picoplankton communities in many marine ecosystems. However, little is known about their community composition in the subtropical coastal waters of the Northwestern Pacific Ocean. In order to study their taxonomic composition, this study constructed 18S rRNA gene libraries using flow cytometric sorting during the warm season. The results show that, after diatoms, prasinophyte clones are numerically dominant. Within prasinophytes, Micromonas produced the most common sequences, and included clades II, III, IV, and VI. We are establishing the new Micromonas clade VI based on our phylogenetic analysis. Sequences of this clade have previously been retrieved from the South China Sea and Red Sea, indicating a worldwide distribution, but this is the first study to detect clade VI in the coastal waters of Taiwan. The TSA-FISH results indicated that Micromonas clade VI peaked in the summer (~4 × 102  cells/ml), accounting for one-fifth of Micromonas abundance on average. Overall, Micromonas contributed half of Mamiellophyceae abundance, while Mamiellophyceae contributed 40% of PPE abundance. This study demonstrates the importance of Micromonas within the Mamiellophyceae in a subtropical coastal ecosystem.