Open-vocabulary segmentation aims to identify and segment specific regions objects based on text-based descriptions. A common solution is leverage powerful vision-language models (VLMs), such as CLIP, bridge the gap between vision text information. However, VLMs are typically pretrained for image-level vision-text alignment, focusing global semantic features. In contrast, tasks require fine-grained pixel-level alignment detailed category boundary information, which alone cannot provide. As a...

To detect fraudulent TLS server certificates and improve the accountability of certification authorities (CAs), certificate transparency (CT) is proposed to record in publicly-visible logs, from which monitors fetch all watch for suspicious ones. However, if monitors, either domain owners themselves or third-party services, fail return a complete set issued interest, potentially may not be detected then CT framework becomes less reliable. This paper presents first systematic study on...

In view of the limitations traditional statistical methods in dealing with multifactor and nonlinear data inadequacy classical machine learning algorithms predicting high dimensions large sample sizes, this paper proposes an operational risk prediction model based on automatic encoder convolutional neural networks. First, we use to extract features motion factors obtain feature components that can highly represent risk. Secondly, causal relationship between sports characteristics, a network...

Multimodal semantic segmentation shows significant potential for enhancing accuracy in complex scenes. However, current methods often incorporate specialized feature fusion modules tailored to specific modalities, thereby restricting input flexibility and increasing the number of training parameters. To address these challenges, we propose StitchFusion, a straightforward yet effective modal framework that integrates large-scale pre-trained models directly as encoders fusers. This approach...

Multimodal semantic segmentation is a pivotal component of computer vision and typically surpasses unimodal methods by utilizing rich information set from various sources.Current models frequently adopt modality-specific frameworks that inherently biases toward certain modalities. Although these might be advantageous in specific situations, they generally limit the adaptability across different multimodal contexts, thereby potentially impairing performance. To address this issue, we leverage...

One of the most fundamental issues in computer security is protecting sensitive files from unauthorized access. Traditional file protection tools run inside target operating system, which hosts files. This makes previous approaches vulnerable face a compromised OS. To address this limitation, recent seek for good isolation by putting system into dedicated virtual machine or using network system. However, they suffer sharp increase trusted computing base size degrades their reliability.

In order to enhance the accountability of certification authorities (CAs), certificate transparency (CT) is proposed record CA-signed certificates in public logs. A accepted by CT-compliant browsers only if it recorded logs, so that any fraudulent will be detected domain owner public-visible practice, third-party monitors fetch logs provide search services, and regularly searches all issued for its from watch suspicious ones among them. this paper, we study links CT framework CAs, browsers,...

Focusing on obtaining in-memory evidence, current live acquisition efforts either fail to provide accurate native system physical memory at the given time point or require suspending machine and altering execution environment drastically. To address this issue, we propose Vis, a light-weight virtualization approach retrieving of content while preserving target system. Vis encapsulates into single virtual then conducts by manipulating nested page table in hypervisor. We present design...

To validate a certificate, user needs to install the certificate of root certification authority (CA) and download revocation information (CRI). Although operating systems browsers manage trust list (CTL) publicly-trusted CAs for global users, locally-trusted still play an important role it is difficult its CTL properly by itself. Meanwhile, CRI access inefficient, sometimes even unavailable, causes privacy leakage. We revisit these problems analyzing TLS sessions within organization. best...

Traditional public key infrastructure (PKI) only provides authentication for network communication, and the standard X.509 certificate used in this architecture reveals user's identity. This lack of privacy protection no longer satisfies increasing demands personal privacy. Though an optimized anonymous PKI realizes anonymity, it has potential to be abused due identity tracking. Therefore, maintaining a balance between user anonymity traceability become requirement current PKI. paper...

Text has become the predominant form of communication on social media, embedding a wealth emotional nuances. Consequently, extraction information from text is paramount importance. Despite previous research making some progress, existing sentiment analysis models still face challenges in integrating diverse semantic and lack interpretability. To address these issues, we propose quantum-inspired deep learning architecture that combines fundamental principles quantum mechanics (QM principles)...

The certificate transparency (CT) framework has been deployed to improve the accountability of TLS ecosystem.However, current implementation CT does not enforce or guarantee correct behavior third-party monitors, which are essential components framework, and raises security reliability concerns.For example, recent studies [32], [33] reported that 5 popular monitors cannot always return complete set certificates inquired by users, fundamentally impairs protection aims offer.This work revisits...