Estimating post-click conversion rate (CVR) accurately is crucial for ranking systems in industrial applications such as recommendation and advertising. Conventional CVR modeling applies popular deep learning methods achieves state-of-the-art performance. However it encounters several task-specific problems practice, making challenging. For example, conventional models are trained with samples of clicked impressions while utilized to make inference on the entire space all impressions. This...

Nowadays, mobile devices are ubiquitous in people's everyday life and applications on becoming increasingly resource-hungry. However, the resources limited. Mobile cloud computing addresses resource scarcity problem of by offloading computation and/or data from into cloud. In converging progress computing, cloudlet is an important complement to client-cloud hierarchy. This paper presents extensive survey researches based computing. We first retrospect evolution After that, we review existing...

Biometric-based authentication is gaining increasing attention for wearables and mobile applications. Meanwhile, the growing adoption of sensors in also provides opportunities to capture novel wearable biometrics. In this work, we propose EarDynamic, an ear canal deformation based user using in-ear wearables. EarDynamic continuous passive transparent users. It leverages that combines unique static geometry dynamic motions when speaking authentication. utilizes acoustic sensing approach with...

Searchable encryption allows mobile devices with limited computing and storage resources to outsource data an untrusted cloud server. Users are able search retrieve the outsourced data; however, it suffers from information privacy leakage. The reason is that most of previous works rely on single model, which server gets all users. In this paper, we present a new scheme M-SSE achieves both forward backward security based multi-cloud technique. secure against adaptive file injection attack...

Monitoring Earth's evolving land covers requires methods capable of detecting changes across a wide range categories and contexts. Existing change detection are hindered by their dependency on predefined classes, reducing effectiveness in open-world applications. To address this issue, we introduce open-vocabulary (OVCD), novel task that bridges vision language to detect any category. Considering the lack high-quality data annotation, propose two training-free frameworks, M-C-I I-M-C, which...

Intel SGX is a hardware-based trusted execution environment (TEE), which enables an application to compute on confidential data in secure enclave. assumes powerful threat model, only the CPU itself trusted; anything else untrusted, including memory, firmware, system software, etc. An enclave interacts with its host through exposed, enclave-specific, (usually) bi-directional interface. This interface main attack surface of The attacker can invoke any order and inputs. It thus imperative it...

Machine learning algorithms are widely used for cybersecurity applications, include spam, malware detection. In these the machine model has to face attack by adversarial samples. Therefore, how train a robust with small samples is very hot research problem. portable document format (PDF) file format, and often utilized as vehicle malicious behavior. There have been various PDF detectors based on learning. However, labeling of large-scale data time-consuming laborious. This paper aims reduce...

In consideration of feasibility, searchable encryption schemes in multi-user setting have to handle the problem dynamical user injection and revocation, especially make sure that revocation will not cause security issues, such as secret key leakage. Recently, fine-grained access control using trusted third party is proposed resolve this issue, however, it increases management complexity for maintaining massive authentication information users. paper, we first time present new concept...

Existing investigations of opponent modeling and intention inferencing cannot make clear descriptions practical explanations the opponent's behaviors intentions, which may inevitably limit applicability them. In this work, we propose a novel approach for policy explanation inference based on behavioral portrait opponent. Specifically, use multiagent deep deterministic gradients (MADDPG) algorithm to train agent in competitive environment, collect data agent's observations. Then perform...

Users of Android phones increasingly entrust personal information to third-party apps. However, recent studies reveal that many apps, even benign ones, could leak sensitive without user awareness or consent. Previous solutions either require modify the framework thus significantly impairing their practical deployment, be easily defeated by malicious apps using a native library.

Response process data collected from human-computer interactive items contain detailed information about respondents' behavioural patterns and cognitive processes. Such are valuable sources for analysing problem-solving strategies. However, the irregular format complex structure make standard statistical tools difficult to apply. This article develops a computationally efficient method exploratory analysis of such data. The new approach segments lengthy individual into sequence short...

Nowadays, the object detection techniques have been developed rapidly for different applications, ranging from remote sensing to autonomous vehicles. We demonstrate identification of open angle and direction using machine learning (ML) algorithms based on received light beam's intensity profiles. Compared with previous optical orbital-angular-momentum (OAM) spectrum system other related works, our proposed technique only uses a single-shot image, can efficiently reduce complexity hardware...

Most existing image restoration methods use neural networks to learn strong image-level priors from huge data estimate the lost information. However, these works still struggle in cases when images have severe information deficits. Introducing external or using reference provide also limitations application domain. In contrast, text input is more readily available and provides with higher flexibility. this work, we design an effective framework that allows user control process of degraded...

Android allows apps to communicate with its system services via service helpers so that these can use various functions provided by the services. Meanwhile, rely on their enforce security checks for protection. Unfortunately, in may be bypassed directly exploiting non-SDK (hidden) APIs, degrading stability and posing severe threats such as privilege escalation, automatic function execution without users' interactions, crashes, DoS attacks. Google has proposed approaches address this problem,...

Acetylcholinesterase has long been considered as a target for Alzheimer disease therapy. In this work, several classification models were built the purpose of distinguishing acetylcholinesterase inhibitors (AChEIs) and decoys. Each molecule was initially represented by 211 ADRIANA.Code 334 MOE descriptors. Correlation analysis, F-score attribute selection methods in Weka used to find best reduced set descriptors, respectively. Additionally, using Support Vector Machine evaluated 5-, 10-fold...

Nowadays, machine learning is widely used in malware detection system as a core component. The algorithm designed under the assumption that all datasets follow same underlying data distribution. But real-world distribution not stable and changes with time. By exploiting knowledge of concept drift problem, we show novel evasive botnet architecture stealthy secure C&C mechanism. Based on email communication channel, construct email-based P2P-like exploit excellent reputation servers huge...

Many machine-learning-based intrusion detection methods have been proposed, however there is a lack of collaboration among these methods. Faced with cascade malicious behaviors and various running environments, coupled the endless emergence new activities, it difficult for us to choose an algorithm manually that suitable all scenarios. In addition, usually binary models are applied only "normal" or "abnormal" decision made, know how much confidence we in prediction model. this study, propose...