Pre-training based deep learning has been used for improving the accuracy of malicious traffic detection. When users use pre-trained encoders to avoid pre-training cost, they may suffer from backdoor attack. In other words, attackers can inject backdoors into these encoders, and trigger backdoors to evade malicious traffic detection. Thus it is necessary to detect the backdoors of these encoders before they are deployed. However, it is challenging, because: (1) traffic triggers are generated dynamically, (2) and encoders output embedding vectors. We hope to use the explanation collapse phenomenon (i.e., backdoor injection changes explanation results) to overcome these challenges. However, existing explanation algorithms are not suitable, because: (1) encoders output embedding vectors, (2) and encoders have no detection boundary. Thus it is necessary to design a special explanation algorithm. We propose MEB, a backdoor detection framework for pre-training based malicious traffic detection. In the detection framework of MEB, we use detection datasets and the explanation algorithm to calculate detection features, and input them into meta classifiers. We detect backdoors according to the prediction results of meta classifiers. In the explanation algorithm of MEB, we calculate the distance between the embedding vectors of explained samples and the embedding vector centroids of benign/malicious samples. We use the gradients of explained samples to the distance to calculate the explanation results. We implement and evaluate MEB in two typical malicious traffic datasets and two typical pre-training algorithms. Experiment results show that MEB can achieve higher detection accuracy and efficiency than baseline detection algorithms.

ABSTRACT In open‐network environments, two‐area load frequency control (LFC) systems are exposed to potential threats from false data injection attacks (FDIAs). Conventional model‐based detection and control methods exhibit poor adaptability to unstructured disturbances, making it difficult to ensure system stability and robustness. To address this issue, a two‐tier defense mechanism is proposed, in which both the front‐end detection and the back‐end control components are designed in a model‐free manner. The detection module adopts recursive estimation and model‐free disturbance observation, while the control module employs a feedback optimal bounded error learning (FOBEL) strategy built on reinforcement learning. The detection module identifies attacks through state residual analysis and signal disturbance estimation, while the control module implements dynamic compensation using a controller that integrates fractional‐order structures with reinforcement learning. Compared with traditional methods, this approach demonstrates significant improvements in disturbance rejection and control accuracy. Simulation studies under two representative attack scenarios validate the superiority and effectiveness of the proposed method in terms of frequency deviation suppression, power fluctuation mitigation, and estimation accuracy.

Semantic segmentation is a key technique involved in automatic interpretation of high-resolution remote sensing (HRS) imagery and has drawn much attention in the remote sensing community. Deep convolutional neural networks (DCNNs) have been successfully applied to the HRS imagery semantic segmentation task due to their hierarchical representation ability. However, the heavy dependence on a large number of training data with dense annotation and the sensitiveness to the variation of data distribution severely restrict the potential application of DCNNs for the semantic segmentation of HRS imagery. This study proposes a novel unsupervised domain adaptation semantic segmentation network (MemoryAdaptNet) for the semantic segmentation of HRS imagery. MemoryAdaptNet constructs an output space adversarial learning scheme to bridge the domain distribution discrepancy between the source domain and the target domain and to narrow the influence of domain shift. Specifically, we embed an invariant feature memory module to store invariant domain-level prototype information because the features obtained from adversarial learning only tend to represent the variant feature of current limited inputs. This module is integrated by a category attention-driven invariant domain-level memory aggregation module to current pseudo-invariant feature for further augmenting the representations. An entropy-based pseudo label filtering strategy is used to update the memory module with high-confident pseudo-invariant feature of current target images. Extensive experiments under three cross-domain tasks indicate that our proposed MemoryAdaptNet is remarkably superior to the state-of-the-art methods. Our code is available athttps://github.com/RS-CSU/MemoryAdaptNet-master.

In this paper, we consider a long-distance point-to-point communication system with only a single buffer in which the source generates status updates with rate λ and can only transmit one update at a time to the receiver. The timeliness of the status updates is evaluated by the age of information (AoI). In this setting, two scheduling policies, namely preemption and non-preemption respectively, are adopted to minimize the AoI. Specifically, we investigate the priority of the two scheduling policies in the presence of non-trivial propagation delay, which has received little attention in the existing work. Utilizing the evolution of AoI, explicit expressions of the limiting average age for the two scheduling policies are derived, based on which we theoretically prove that for given λ, there exists a threshold of the propagation delay, within which preemption policy outperforms non-preemption policy from the perspective of the limiting average age. We further formulate an optimization problem minimizing the limitng average age under the constraint of decoding failure probability for the two scheduling policies and determine the optimal codeword length. Numerical results are provided to validate our theoretical analysis.

In this paper, we study the timely transmission of status updates from multiple sources to an interested receiver. The sources generate updates independently at Poisson rate, which are then coded and transmitted through a shared block fading channel. Age of information (AoI) is an effective indicator to describe the timeliness of transmission by depicting the freshness of information the receiver knows about the interested source. An intuitively age-oriented transmission scheme is sequentially independent transmission, i.e., the transmitter submits the updates one-by-one without waiting. However, this scheme does not always minimize the age. Aiming to improve the age performance of the considered system, we propose a waiting-and-batching transmission scheme. Specifically, earlier generated updates are waiting for late ones and then batching into a long packet before being coded and transmitted. We derive the close-formed expression of the average AoI of the proposed scheme, and conduct extensive numerical comparisons with the independent transmission scheme. Results reveal that our waiting-and-batching scheme can be more beneficial to AoI if: (1) the update generation rate is high; (2) the propagation delay is non-trivial.

Tunnel FET (TFET) is recognized to be one of the most promising candidates for ultra-low power applications due to its ultra-low off current and high compatibility with CMOS process. However, different from the typical features of MOSFET, some electrical characteristics of TFETs caused by asymmetric device structure and special conduction mechanism may make conventional topologies of circuits no longer applicable. In this paper, it is found that the TFETs stacking will result in severe current degradation behavior, which makes traditional topologies of logic gates may be not applicable. To solve this problem, a set of novel hybrid TFET-MOSFET topologies for standard logic cells are proposed. The proposed designs achieve more than 2 times lower hardware cost and intrinsic delay, and realize up to 4 times lower area-power-delay product (APDP) than that of conventional TFET-based logic circuits. Moreover, the proposed topologies can achieve almost 2 orders of magnitude lower power and up to 34 times lower APDP than that of conventional MOSFET-based logic circuits. The proposed standard logic cells show great superiority for power-constraint applications.

Event mining is a useful way to understand computer system behaviors. The focus of recent works on event mining has been shifted to event summarization from discovering frequent patterns. Event summarization seeks to provide a comprehensible explanation of the event sequence on certain aspects. Previous methods have several limitations such as high time complexity, a low precision especially with the presence of noise and phase shifts, and providing a summary which is difficult for a human to understand. In this paper, we propose an integrated event summarization approach toward the understanding of the chaotic temporal data. Our approach focuses on two kinds of temporal relationships, the periodic pattern and the correlation pattern, hidden in event sequences of at most two event types. For the periodic patterns, we propose an event periodicity detection algorithm to discover them directly. For the correlation patterns, we make a simple statistical test based on low order statistics to check temporal dependency of events of two types and to eliminate event correlation candidate space dramatically, and then apply inter-arrival histograms to summarize an event sequence and capture the correlation patterns. In order to balance between accuracy and brevity, the minimum description length principle is used to guide the summarization process. Further, the event relationship network is built to describe discovered patterns. We conduct several groups of experiments on synthetic and real data. Experimental results show our approach is capable of producing usable event summarization, robust to noises, and scalable. The average compression ratio of event sequences reaches 99.7%.

In this paper, a novel combinational access topology of Tunnel FET (TFET) SRAM is proposed for ultra-Low Power applications. Since forward p-i-n current of TFET could cause serious damage to SRAM circuit performance, the proposed topology can avoid the forward bias applied to the p-i-n junction, thus increasing SRAM cell read and hold static noise margin (SNM) and decreasing its static power consumption dramatically. At 0.6 V supply voltage, the combinational access TFET SRAM topology presents 26% hold SNM larger than traditional TFET SRAM topologies, 8 orders of magnitude lower static power consumption, and 2 order of magnitude lower power delay product, demonstrating its great potential for ultra-low power applications.

Low power applications have led to a boom in researches on new circuits based on steep-slope transistors, of which the objective is to overcome MOSFET's drawback of inevitable increasing leakage power while maintaining acceptable performance in low voltage operation. Among those emerging transistors, Tunnel FET (TFET) becomes a most promising one due to its low off current and compatibility with CMOS process. In order to guide the application and the improvement of TFET, in this paper from a circuit-level perspective, utilizing a newly defined benchmarking method, we figured out the frequency-VDD range in which Si TFET circuits show low power advantage over their MOSFET counterparts based on HSPICE simulations using calibrated compact model. A systematic and quantitative analysis was then conducted to further enlarge the application scope of TFET circuits, with a Figure of Merit (FOM) and a guideline for future TFET proposed.

Noisy video occurs frequently in many visual communication applications, especially in video surveillance and video phone. The noise affects compression a lot by decreasing the accuracy of prediction and increasing the residual. In this paper, we propose a novel inter-frame filtering method based on motion compensation for reference picture in video coding. It constructs the low-pass temporal filter by modulating the constructed picture with previous reference picture. As a result, it greatly reduces the noise in reference picture, which results in the improvement of compression efficiency. As the replacement of traditional reference picture, this proposed dynamical motion compensation based Synthesized Reference (SR) picture is computation efficient and easy for implementation. It can be applied to any block-based hybrid coding structure, such as H.264/AVC, AVS (Audio Video Coding Standard), MPEG-4. Simulation results show that more than 20% bit rate is reduced when applying SR picture to the encoding of typical video surveillance sequences. At the same time, the visual quality is greatly improved.

In an H.264 video encoder, motion estimation (ME) is the most time-consuming component. The ME process consists of two stages, integer pixel search and fractional pixel search. Since the complexity of integer pixel search has been greatly reduced by numerous fast ME algorithms, the computation overhead required by fractional pixel ME has become relatively significant. To reduce the complexity of fractional pixel ME, we propose a prediction-based directional fractional pixel ME algorithm. We utilize more accurate motion vector predictions and directional search to achieve better computation reduction. We further propose an early termination method to decrease the amount of search. Experimental results show that, compared to the full search sub-pel ME and the fast sub-pel ME proposed in H.264, the proposed method can reduce up to 84% and 74% of fractional pixel search points respectively, with a negligible degradation in quality.

One of UWB proposal in IEEE P802.15 WPAN project is to use OFDM system and punctured convolutional codes for UWB channels supporting data rate up to 480 Mb/s. We apply a TTCM encoder to examine the possible improvement for UWB/OFDM. We construct punctured parity-concatenated trellis codes in which a TCM code is used as the inner code and a simple parity-check code is used as the outer code. The bit MAP algorithm is used in iterative decoding. The code performance is examined when applied to OFDM systems in UWB channel environments. The study shows that the system can offer much higher spectral efficiency up to 1 Gbps over UWB channel

The H.264 video coding standard provides considerably higher coding efficiency than previous standards do, whereas its complexity is significantly increased at the same time. In an H.264 encoder, the most time-consuming component is variable block-size motion estimation. To reduce the complexity of motion estimation, an early termination algorithm is proposed in this paper. It predicts the best motion vector by examining only one search point. With the proposed method, some of the motion searches can be stopped early, and then a large number of search points can be skipped. The proposed method can work with any fast motion estimation algorithm. Experiments are carried out with a fast motion estimation algorithm that has been adopted by H.264. Results show that significant complexity reduction is achieved while the degradation in video quality is negligible.

This paper presents a sampling-based approach to computing and executing feedback-motion strategies by defining a global navigation function over a collection of neighborhoods in configuration space. The collection of neighborhoods and their underlying connectivity structure are captured by a sampling-based neighborhood graph (SNG), on which navigation functions are built. The SNG construction algorithm incrementally places new neighborhoods in the configuration space, using distance information provided by existing collision-detection algorithms. A termination condition indicates the probability that a specified fraction of the space is covered. Our implementation illustrates the approach for rigid and articulated bodies with up to six-dimensional configuration spaces. Even over such spaces, rapid online responses to unpredictable configuration changes can be made in a few microseconds on standard PC hardware. Furthermore, if the goal is changed, an updated navigation function can be quickly computed without performing additional collision checking.

As a general framework to determine a collision-free feedback motion strategies, the Random Neighborhood Graph (RNG) approach [19] defines a global navigation function over an approximate representation of the free configuration. In this paper, we improve the RNG approach in several aspects. We present an ANN-accelerated RNG construction a1gorithm to achieve near 1ogarithmic running time in each iteration of the RNG expansion. Two probabilistic termination conditions of the RNG constructibn a1gorithm are presented and analyzed. To help overcome the difficulty of narrow corridors, we also introduce a randoniized perturbation algorithm to enhance the sampling quality. Our implementation illustrates a significant performance improvement.

We propose the use of quasi-random sampling techniques for path planning in high-dimensional configuration spaces. Following similar trends from related numerical computation fields, we show several advantages offered by these techniques in comparison to random sampling. Our ideas are evaluated in the context of the probabilistic roadmap (PRM) framework. Two quasi-random variants of PRM- based planners are proposed: 1) a classical PRM with quasi-random sampling; and 2) a quasi-random lazy-PRM. Both have been implemented, and are shown through experiments to offer some performance advantages in comparison to their randomized counterparts.

This paper presents a randomized framework for computing feedback motion strategies, by defining a global navigation function over a collection of spherical balls in the configuration space. If the goal is changed, an updated navigation function can be quickly computed, offering benefits similar to the fast multiple queries permitted by the probabilistic roadmap approach to path planning. Our choice of balls is motivated in part by recent tools from computational geometry which compute point locations and arrangements efficiently without significant dependence on dimension. We present a construction algorithm that includes a Bayesian termination condition based on the probability that a specified fraction of the free space is covered. A basic implementation illustrates the framework for rigid and articulated bodies with up to five-dimensional configuration spaces.