Network-based Intrusion Detection System (NIDS) forms the frontline defence against network attacks that compromise security of data, systems, and networks. In recent years, Deep Neural Networks (DNNs) have been increasingly used in NIDS to detect malicious traffic due their high detection accuracy. However, DNNs are vulnerable adversarial modify an input example with imperceivable perturbation, which causes a misclassification by DNN. security-sensitive domains, such as NIDS, pose severe...

IP traffic classification has been a vitally important topic that attracts persistent interest in the networking and machine learning communities for past decades. While there exist quite number of works applying techniques to realize classification, most suffer from limitations like either heavily depending on handcrafted features or be only able handle offline classification. To get rid aforementioned weakness, this paper, we propose our online Convolutional Neural Networks (CNNs) based...

Driven by economic benefits, the number of malware attacks is increasing significantly on a daily basis. Malware Detection Systems (MDS) first line defense against malicious attacks, thus it important for detection systems to accurately and efficiently detect malware. Traditional MDS typically utilizes traditional machine learning algorithms that require feature selection extraction, which are time-consuming error-prone. Conventional deep based approaches use Recurrent Neural Network (RNN)...

SUMMARY Software security issues have been a major concern in the cyberspace community, so great deal of research on testing has performed, and various techniques developed. Threat modeling provides systematic way to identify threats that might compromise security, it well‐accepted practice by industry, but test case generation from threat models not addressed yet. Thus, this paper, we propose model‐based approach automatically generates sequences trees transforms them into executable tests....

With the rapid advancement of ocean monitoring technology, types and quantities underwater sensors have increased significantly. Traditional single-sensor approaches exhibit limitations in target classification, resulting low classification accuracy poor robustness. This paper integrates deep learning information fusion theory to propose a multi-level perception method for targets based on multi-physical-field sensing. We extract both conventional typical features derived from an autoencoder...

Network Intrusion Detection Systems (NIDSes) are crucial for securing various networks from malicious attacks. Recent developments in Deep Neural Networks (DNNs) have encouraged researchers to incorporate DNNs as the underlying detection engine NIDS. However, susceptible adversarial attacks, where subtle modifications input data result misclassification, posing a significant threat security-sensitive domains such Existing efforts defenses predominantly focus on supervised classification...

Memory resources in data centers generally suffer from low utilization and lack of dynamics. disaggregation solves these problems by decoupling CPU memory, which currently includes approaches based on RDMA or interconnection protocols such as Compute Express Link (CXL). However, the RDMA-based approach involves code refactoring higher latency. The CXL-based supports native memory semantics overcomes shortcomings RDMA, but is limited within rack level. In addition, pooling sharing CXL...

Software security issues have been a major concern to the cyberspace community, so great deal of research on testing has performed, and various techniques developed. Most these techniques, however, focused software systems after their implementation is completed. To build secure dependable in cost-effective way, it necessary put more effort upfront during development life cycle. In this paper, we provided approach that derives test cases from design-level artifacts. The consider consists...

This paper presents a unified threat model for assessing in web applications. We extend the tree with more semantic and context information about to form new which is used analyze evaluate software design stage. utilize historical statistical contained this mitigation schemes. The results schemes can be direct secure coding testing. makes it possible threat-resistant applications by means of detecting mitigating early

This paper presents an attack scenario based approach for software security testing at design stage. Attack scenarios are represented as extended activity diagram (EAD) and new unified threat model (NUTM). Security test cases derived from automatically according to coverage criteria of complex path. These applied the system. According case results, system can be improved by mitigations. In addition, pattern provided developers characterize reuse well-studied attacks mitigations in a quick...

Deep learning has becoming increasingly more popular in recent years, and there are many frameworks the market accordingly, such as Caffe, TensorFlow Pytorch. All these natively support CPUs GPGPUs. However, FPGAs still cannot provide a comprehensive by for deep development, especially training phase. In this paper, we firstly propose FeCaffe, i.e. FPGA-enabled hierarchical software hardware design methodology based on to enable FPGA CNN features. Moreover, some benchmarks of networks with...

Abstract Wear is one of the major causes that affect performance and reliability tribo-systems. To mitigate its adverse effects, it necessary to monitor wear progress so preventive maintenance can be timely scheduled. An online visual ferrograph (OLVF) apparatus used obtain measurements particle quantities, wearing a four-ball tribometer under different lubrication conditions, several popular deep learning algorithms are evaluated for their effectiveness in providing decisions. The obtained...

In this paper, a constrained bi-RRT*-smart algorithm is proposed to solve the recovery path planning problem of an Autonomous Underwater Vehicle (AUV). The RRT*-smart taken as basic strategy for predefined path-planning generate optimal which has smallest distance and no collisions with islands. On basis, bi-direction technique fusion get bi- algorithm. Additionally, two practical constraints are considered facilitate feasibility method in real applications. For one thing, considering...

ABSTRACT Design‐level vulnerabilities are a major source of security problems in software programs. For the purpose improving trustworthiness designs, this paper presents unified threat model for representing, analyzing, and evaluating threats at various design stages. Unified models represent via tree structures with AND/OR logical relationships evaluates cost‐effective way based on attack paths. Mitigation measures designed prioritized evaluation results, which make it possible to...

In this paper, a stagewise approach has been adopted in the Direction of Arrival (DOA) estimation procedure to improve speed and efficiency whole estimation, i.e. coarse is firstly carried out by means Minimum Variance Distortionless Response (MVDR) provide smaller searching angle for secondary high precision estimator, which realized modified Orthogonal Matching Pursuit (OMP) algorithm with access spatial over-complete dictionary. Simulations demonstrate that compared methods are based on...

Finger motion recognition is one of the key technologies human-computer interaction based on gestures.In this paper, we propose a method recognizing finger motions by using wearable wrist device (WWD).This not only avoids problem that user's hands are limited wearing detection sensors, but also vision-sensorbased gesture technology difficult to use in mobile environment.Moreover, method, which uses polyvinylidene fluoride (PVDF) sensors as units WWD, has advantages being noninvasive,...

Embedding as a Service (EaaS) has become widely adopted solution, which offers feature extraction capabilities for addressing various downstream tasks in Natural Language Processing (NLP). Prior studies have shown that EaaS can be prone to model attacks; nevertheless, this concern could mitigated by adding backdoor watermarks the text embeddings and subsequently verifying attack models post-publication. Through analysis of recent watermarking strategy EaaS, EmbMarker, we design novel CSE...

The use of multiple features for tracking has been proved as an effective approach, the reason that limitation each feature could be compensated. Sparse representation with a particle filter is one most influential frameworks visual tracking. Many sparse algorithms consider holistic or local target, and rarely take into account distribution coefficient. In this paper, we propose group model using object By appropriately selecting to form dictionary, based scheme can assign proper weights...

Due to its high expressiveness and speed, Deep Learning (DL) has become an increasingly popular choice as the detection algorithm for Network-based Intrusion Detection Systems (NIDSes). Unfortunately, DL algorithms are vulnerable adversarial examples that inject imperceptible modifications input cause misclassify input. Existing attacks in NIDS domain often manipulate traffic features directly, which hold no practical significance because cannot be replayed a real network. It remains...