Trace of the Times: Rootkit Detection through Temporal Anomalies in Kernel Activity

Open soft real-time systems, such as mobile robots, experience unpredictable interactions with their environments and yet must respond both adaptively and with reasonable temporal predictability. Because of the uncertainty inherent in such interactions, many of the assumptions of the real-time scheduling techniques traditionally used to ensure predictable timing of system actions do not hold in those environments. In previous work we have developed novel techniques for scheduling policy design where up-front knowledge of execution time distributions can be used to produce both compact representations of resource utilization state spaces and efficient optimal scheduling policies over those state spaces. This paper makes two main contributions beyond our previous work, to the state of the art in scheduling open soft real-time systems: (1) it shows how to relax the assumption that the entire distribution of execution times is known up front, to allow online learning of an execution time distribution during system run-time; and (2) it shows how to relax the assumption that the execution time of a system action can be characterized by a single distribution, to accommodate different execution time distributions for an action being taken in one of multiple modes. Each of these contributions allows a wider range of system actions to be scheduled adaptively and with Notes: On-line version of paper submitted to RTSS 2009, with full proof in Appendix A. Type of Report: Other Department of Computer Science & Engineering Washington University in St. Louis Campus Box 1045 St. Louis, MO 63130 ph: (314) 935-6160 Scheduling Design with Unknown Execution Time Distributions or Modes Robert Glaubius, Terry Tidwell, Christopher Gill, and William D. Smart {rlg1,ttidwell, cdgill, wds}@cse.wustl.edu Department of Computer Science and Engineering Washington University, St. Louis Abstract— Open soft real-time systems, such as mobile robots, experience unpredictable interactions with their environments and yet must respond both adaptively and with reasonable temporal predictability. Because of the uncertainty inherent in such interactions, many of the assumptions of the real-time scheduling techniques traditionally used to ensure predictable timing of system actions do not hold in those environments. In previous work we have developed novel techniques for scheduling policy design where up-front knowledge of execution time distributions can be used to produce both compact representations of resource utilization state spaces and efficient optimal scheduling policies over those state spaces. This paper makes two main contributions beyond our previous work, to the state of the art in scheduling open soft realtime systems: (1) it shows how to relax the assumption that the entire distribution of execution times is known up front, to allow online learning of an execution time distribution during system run-time; and (2) it shows how to relax the assumption that the execution time of a system action can be characterized by a single distribution, to accommodate different execution time distributions for an action being taken in one of multiple modes. Each of these contributions allows a wider range of system actions to be scheduled adaptively and with temporal predictability, which increases the applicability of our approach to even more general classes of open soft real-time systems. Open soft real-time systems, such as mobile robots, experience unpredictable interactions with their environments and yet must respond both adaptively and with reasonable temporal predictability. Because of the uncertainty inherent in such interactions, many of the assumptions of the real-time scheduling techniques traditionally used to ensure predictable timing of system actions do not hold in those environments. In previous work we have developed novel techniques for scheduling policy design where up-front knowledge of execution time distributions can be used to produce both compact representations of resource utilization state spaces and efficient optimal scheduling policies over those state spaces. This paper makes two main contributions beyond our previous work, to the state of the art in scheduling open soft realtime systems: (1) it shows how to relax the assumption that the entire distribution of execution times is known up front, to allow online learning of an execution time distribution during system run-time; and (2) it shows how to relax the assumption that the execution time of a system action can be characterized by a single distribution, to accommodate different execution time distributions for an action being taken in one of multiple modes. Each of these contributions allows a wider range of system actions to be scheduled adaptively and with temporal predictability, which increases the applicability of our approach to even more general classes of open soft real-time systems.

Position detection technologies are widely used such as autonomous robot, self-driving, automatic guided vehicles, etc. GPS (Global Positioning System) position measurement is a typical method for position detection, but it is difficult for indoor applications. There are a lot of methods for indoor position detection, for example, magnetic induction method, position detection learning system, and optical position detection system. The magnetic induction method is convenient but needs cost for repairing and route changing. The position detection learning system needs high performance computer and position detection accuracy is low. The optical position detection system has high detection accuracy but it is difficult for optical setting. In addition, the cost of transmitter and receiver becomes expensive for wide area. A new approach of optical position detection system is introduced that using the ceiling LED (Light Emitting Diode) lights and a PSD (Position Sensitive Detector). The general LED lights are sine-waved with different frequencies and the two-dimensional PSD detects the position of LED lights of different frequencies. Using the information of different LED lights, the position of the PSD can be measured. It has high detection accuracy and easy for route changing. Some examples are concerned for indoor applications.

Background Patients with inferior wall myocardial infarction who have right ventricular (RV) involvement appear to have a worse prognosis than those without RV involvement; infarcted RV tissue fails to offer a sufficient preload which is essential for adequate LV performance. Thus, assessment of RV function is an important step in dealing with patients presenting with inferior wall myocardial infarction that will help in adopting a proper management plan. Objective To assess the correlation between RV function and angiographic findings in patients presenting with inferior wall myocardial infarction associated with RV infarction undergoing primary percutaneous coronary intervention. Patients and Methods Study included 60 patients who presented to Ain shams university hospitals by inferior wall ST segment elevation myocardial infarction associated with RV infarction during the period from February 2019 to August 2019.All patients were subjected to history taking, clinical examination, ECG recording then primary percutaneous coronary intervention. Echocardiographic assessment was done to all patients within 48 hours of admission. Results Study included 60 patients, 43 males (71.7%) and 17 females (28.3%), with mean age of 56.73 ± 9.94 years. Commonest Infarction related Artery (IRA) associated with impaired RV function was proximal RCA (p-value: 0.003). In 23 patients (38.3%) heavy thrombus burden was found while in the other 37 patients (61.7%) there was no evidence of heavy thrombus burden. Regarding post procedural TIMI flow grade: 1 patient (1.7%) had final TIMI I flow, 9 patients (15.0%) had final TIMI II flow and 50 patients (83.3%) had final TIMI III flow. There was statistically significant relationship between RV function assessed through measuring RV free wall strain and both of thrombus burden and final TIMI flow grade. Abnormal RV function was more commonly associated with heavy thrombus burden (p-value:0.023) and less than TIMI III flow after angioplasty (p-value:0.011).RV free wall systolic strain assessment had highest accuracy (75%) in detection of proximal RCA occlusion compared to other parameters including TAPSE, S’ and FAC. Conclusion Impaired RV function in patients presenting with RV infarction can be predicted by different angiographic findings. Proximal RCA total occlusion being commonest IRA associated with impaired RV function. Also, presence of heavy thrombus burden and less than TIMI III flow after angioplasty are associated with increased risk of impaired RV function. RV free wall strain measured by 2D-speckle tracking echocardiography has highest accuracy in detection of proximal RCA occlusion compared to other echocardiographic indices including TAPSE, S’ and FAC.

Simple SummaryFlow cytometric immunophenotyping is critical in detecting minimal residual disease (MRD) in patients with chronic lymphocytic leukemia (CLL). However, flow cytometric analysis is complicated and time-consuming. Herein, we evaluated the performance of a deep neural network (DNN) in detecting CLL MRD and whether it could improve the diagnostic workflow in a clinical laboratory setting. Our findings demonstrated that a hybrid DNN approach had high accuracy in detecting CLL MRD; it standardized the gating strategy and dramatically reduced gating time, and it could be fully integrated into the existing clinical laboratory.Flow cytometric (FC) immunophenotyping is critical but time-consuming in diagnosing minimal residual disease (MRD). We evaluated whether human-in-the-loop artificial intelligence (AI) could improve the efficiency of clinical laboratories in detecting MRD in chronic lymphocytic leukemia (CLL). We developed deep neural networks (DNN) that were trained on a 10-color CLL MRD panel from treated CLL patients, including DNN trained on the full cohort of 202 patients (F-DNN) and DNN trained on 138 patients with low-event cases (MRD < 1000 events) (L-DNN). A hybrid DNN approach was utilized, with F-DNN and L-DNN applied sequentially to cases. “Ground truth” classification of CLL MRD was confirmed by expert analysis. The hybrid DNN approach demonstrated an overall accuracy of 97.1% (95% CI: 84.7–99.9%) in an independent cohort of 34 unknown samples. When CLL cells were reported as a percentage of total white blood cells, there was excellent correlation between the DNN and expert analysis [r > 0.999; Passing–Bablok slope = 0.997 (95% CI: 0.988–0.999) and intercept = 0.001 (95% CI: 0.000–0.001)]. Gating time was dramatically reduced to 12 s/case by DNN from 15 min/case by the manual process. The proposed DNN demonstrated high accuracy in CLL MRD detection and significantly improved workflow efficiency. Additional clinical validation is needed before it can be fully integrated into the existing clinical laboratory practice.

High detection accuracy in piezoelectric-based force sensing in interactive displays has gained global attention. To achieve this, artificial neural networks (ANN)—successful and widely used machine learning algorithms—have been demonstrated to be potentially powerful tools, providing acceptable location detection accuracy of 95.2% and force level recognition of 93.3% in a previous study. While these values might be acceptable for conventional operations, e.g., opening a folder, they must be boosted for applications where intensive operations are performed. Furthermore, the relatively high computational cost reported prevents the popularity of ANN-based techniques in conventional artificial intelligence (AI) chip-free end-terminals. In this article, an ANN is designed and optimized for piezoelectric-based touch panels in interactive displays for the first time. The presented technique experimentally allows a conventional smart device to work smoothly with a high detection accuracy of above 97% for both location and force level detection with a low computational cost, thereby advancing the user experience, and serviced by piezoelectric-based touch interfaces in displays.

Vision-based object detection is an essential component of autonomous driving. Because vehicles typically have limited on-board computing resources, a small-sized detection model is required. Simultaneously, high object detection accuracy and real-time inference detection speeds are required to ensure safety while driving. In this paper, an anchor-free lightweight object detector for autonomous driving called ALODAD is proposed. ALODAD incorporates an attention scheme into the lightweight neural network GhostNet and builds an anchor-free detection framework to achieve lower computational costs and provide parameters with high detection accuracy. Specifically, the lightweight backbone neural network integrates a convolutional block attention model that analyzes the valuable features from traffic scene images to generate an accurate bounding box, and then constructs feature pyramids for multi-scale object detection. The proposed method adds an intersection over union (IoU) branch to the decoupled detector to rank the vast number of candidate detections accurately. To increase the data diversity, data augmentation was used during training. Extensive experiments based on benchmarks demonstrate that the proposed method offers improved performance compared to the baseline. The proposed method can achieve an increased detection accuracy while meeting the real-time requirements of autonomous driving. The proposed method was compared with the YOLOv5 and RetinaNet models and 98.7% and 94.5% were obtained for the average precision metrics AP50 and AP75, respectively, on the BCTSDB dataset.

Quantification of myocardial perfusion has the potential to improve the detection of regional and global flow reduction. Significant effort has been made to automate the workflow, where one essential step is the arterial input function (AIF) extraction. Failure to accurately identify the left ventricle (LV) prevents AIF estimation required for quantification, therefore high detection accuracy is required. This study presents a robust LV detection method using the convolutional neural network (CNN). CNN models were trained by assembling 25,027 scans (N = 12,984 patients) from three hospitals, seven scanners. Performance was evaluated using a hold-out test set of 5721 scans (N = 2805 patients). Model inputs were a time series of AIF images (2D+T). Two variations were investigated: (1) two classes (2CS) for background and foreground (LV mask), and (2) three classes (3CS) for background, LV, and RV. The final model was deployed on MRI scanners using the Gadgetron reconstruction software framework. Model loading on the MRI scanner took ~340ms and applying the model took ~180ms. The 3CS model successfully detected the LV in 99.98% of all test cases (1 failure out of 5721). The mean Dice ratio for 3CS was 0.87 ± 0.08 with 92.0% of all cases having Dice >0.75. The 2CS model gave a lower Dice ratio of 0.82 ± 0.22 (P < 1e-5). There was no significant difference in foot-time, peak-time, first-pass duration, peak value, and area-under-curve (P > .2) comparing automatically extracted AIF signals with signals from manually drawn contours. A CNN-based solution to detect the LV blood pool from the arterial input function image series was developed, validated, and deployed. A high LV detection accuracy of 99.98% was achieved.

With the proliferation of the availability of highly sophisticated editing tools, the authenticity of digital images has now become questionable. An adversary may perform copy-paste forgery where portions from JPEG image files are copy-pasted into another image file (JPEG, TIFF, etc.) and then save it in uncompressed format (TIFF). It is known that tampered JPEG images contain traces of compression artifacts. In this paper, we propose an automated blind JPEG image forgery detection and localization technique with high detection accuracy, which is effective for identifying traces of JPEG compression in digital images saved in uncompressed format (TIFF). The forgery detection and localization is based on the computation and analysis of a correlation matrix calculated by recompressing the given (possibly tampered) image at different quality factors and then comparing the recompressed versions with the given image. The experimental results prove our technique has high detection and localization accuracy as compared to the existing techniques.

This paper proposes a novel approach for real-time sand detection in gas wells using an Autoencoder and a rule-based model ensemble. Accurate identification and early detection of sand in gas wells are crucial for effective sand management. However, the dynamic flow conditions and high-pressure environments of gas wells make sand detection challenging. The study aims to develop a methodology for sand production detection, considering the limitations of surface facilities and laboratory measurements. Methodology is based on essential features such as Acoustic Sand Detector (ASD) readings, Downhole Pressure, and Production Choke measurements. The methodology was applied for both top-side installed ASDs as well as subsea wells. To overcome the complexity of sand production patterns, we employ an Autoencoder, a deep learning technique capable of addressing the issue. In conjunction, we develop a rule-based model that leverages domain expertise to detect simpler sand patterns. The combination of these models forms a comprehensive and robust sand detection framework for real-time operations and monitoring. The proposed model demonstrates high accuracy in sand detection. After running the model on unsupervised data, we manually evaluated the results by inspecting AI estimated sand labels individually. Labels were assigned to classify them as true positives (captured sands) or false positives (false alarms). Accuracy and precision metrics were then calculated. For the top-side gas wells, the model achieves accuracy ranging from 98.5% to 100%. Meanwhile, for the subsea wells, the model achieves accuracy ranging from 89.4% to 100%. The performance of our approach demonstrates its effectiveness in detecting sand based on recordings both from surface and subsea gas wells. By accurately identifying sand presence, the proposed approach contributes to enhancing sand management strategies in the gas industry. The findings enable proactive maintenance, reduction in equipment damage, and optimization of production processes, ultimately enhancing operational efficiency in gas wells. The algorithm has been successfully deployed for a gas field located in the Caspian Basin. This paper introduces a comprehensive and robust sand detection framework that effectively addresses the complexity of sand production patterns in gas wells. By integrating the Autoencoder with a rule-based model that utilizes domain expertise, creating the framework that demonstrates high accuracy in detecting sand in both topside and subsea wells. This innovative methodology significantly enhances operational efficiency in sand management within the petroleum industry. The findings of this study make a valuable contribution to the literature, offering a novel approach to sand detection that promises to benefit the industry at large.

Current analytic solutions to the execution time distribution of a parallel composition of tasks having stochastic execution times are computationally complex, except for a limited number of distributions. In this paper, we present an analytical solution based on approximating execution time distributions in terms of the first four statistical moments. This low-cost approach allows the parallel execution time distribution to be approximated at ultra-low solution complexity for a wide range of execution time distributions. The accuracy of our method is experimentally evaluated for synthetic distributions as well as for task execution time distributions found in real parallel programs and kernels (NAS-EP, SSSP, APSP, Splash2-Barnes, PSRS, and WATOR). Our experiments show that the prediction error of the mean value of the parallel execution time for N-ary parallel composition is in the order of percents, provided the task execution time distributions are sufficiently independent and unimodal.

Mass measurement is the main outcome of mass spectrometry-based proteomics yet the potential of recent advances in accurate mass measurements remains largely unexploited. There is not even a clear definition of mass accuracy in the proteomics literature, and we identify at least three uses of this term: anecdotal mass accuracy, statistical mass accuracy, and the maximum mass deviation (MMD) allowed in a database search. We suggest using the second of these terms as the generic one. To make the best use of the mass precision offered by modern instruments we propose a series of simple steps involving recalibration of the data on "internal standards" contained in every proteomics data set. Each data set should be accompanied by a plot of mass errors from which the appropriate MMD can be chosen. More advanced uses of high mass accuracy include an MMD that depends on the signal abundance of each peptide. Adapting search engines to high mass accuracy in the MS/MS data is also a high priority. Proper use of high mass accuracy data can make MS-based proteomics one of the most "digital" and accurate post-genomics disciplines.

We address an old but still controversial question of morphological phylogenetics: whether additive (or ordered) coding is beneficial to properly extracting phylogenetic information from phenotypical variation. To empirically evaluate the value of the additive coding, we compared the impact of multistate additive, non-additive, and binary codings for 14 quantitative characters in a phylogenetic analysis of a genus of phorid flies (Diptera). First, we compared which of these morphological codings were most effective for the morphological matrix to approximate the results of a molecular data set. We then compared which morphological coding strategies yielded the best Bayesian posterior probabilities when concatenated to molecular data. We also calculated consistency and retention indices for each binary element of the additive characters and contrasted these results to a measure of phylogenetic signal. Overall, these indices were lower for additive characters than for the others but still indicate reasonable accommodation in the tree. Additive coding outperformed the multistate non-additive coding by recovering higher Bayesian posterior probabilities in the concatenated dataset. Additive coding was also among the best coding strategies for the morphological matrix to approximate the phylogenetic signal from an independent source of evidence—i.e., molecular results. Therefore, quantitative information coded as additive had reasonable phylogenetic congruence with other data and improved the phylogenetic results of morphological data in most cases. These results support the use of additive coding for phylogenetic analysis and encourage other similar empirical evaluations aiming to explore the generality of the benefits of this coding method.

Advances in data mining and machine learning methods for classification and regression open the door of identifying complex patterns from domain sensitive data. In biomedical applications, massive amounts of clinical data are generated and collected to predict diseases. Diagnosis of diseases, such as diabetes and liver diseases, needs more tests these days and that increases the size of patient medical data. Manual exploration of this patient test data is challenging and difficult. Robust prediction of a patient’s liver disease from the massive data set is an important research agenda in health science. The challenge of applying a machine learning method is to select the best algorithm within the disease prediction framework. The goal of this research is to select a robust machine learning algorithm that can equally be applicable on diabetes prediction as well as in liver diseases prediction. This study analyzes two machine learning approaches, support vector machine (SVM) and K-nearest neighbors (KNN) algorithms over two different datasets, diabetes and liver diseases datasets. It was observed that a tuned radial SVM method performed with the highest accuracy in detection of diabetes and liver disease detection with an accuracy of 0.989 for diabetes detection and 0.910 for liver disease detection.

Aim: to work out of a set of measures aimed for early detection of colorectal tumors and the choice of a method of endoscopic surgery.Patients and methods: a multimodal approach was used, which included two successive stages: the stage of assessing the depth of invasion of malignant colorectal epithelial tumors (1) and the stage of endoscopic surgery. The study included 974 patients, aged 67 (43-81) years. The algorithm of the systemic automatic approach to differentiate the depth of invasion of superficial malignant colorectal tumors has been worked out based on analysis of color pictures of colonoscopy (Colonoscopy Video Analysis). The results of use of automatic system were compared with experts’ assessment.Results: the application of the developed algorithm of the systemic automatic approach to differentiate the depth of invasion of malignant ENC has high detection accuracy – the total average detection accuracy when implementing this algorithm is 72.02. No significant differences with experts’ assessment were obtained. With endoscopic removal of malignant tumors with superficial invasion, the correct selection of patients based on the tumor size (up to 2.0 and over 2.0 cm) and the corresponding removal technique (mucosal resection or endoscopic submucosal dissection) are decisive.Conclusion: the automatic system of evaluation of tumor invasion depth has a high accuracy and gives a possibility to exclude false positive results.

Facial expression recognition is a hot research topic in artificial intelligence industry and has a good research prospect in various fields. At present, facial expression recognition processes the whole face directly, but the pixel value of non-feature region may bring some interference for feature descriptor extraction. Considering that the cartoon effect of the face can directly reflect the facial expression features. In order to make the network pay more attention to the information of the facial features and their surrounding pixels, this paper proposes an expression recognition algorithm based on key point detection of Covering multi-scale Practical Landmark Detector (CM-PFLD). Under this algorithm, this paper constructs a cartoon expression data set, which only retains the key points of facial expression information, and then classifies facial expression by directly locating the key points of facial expression information. In order to verify the feasibility of the expression recognition method in this paper. The experiment uses Fer2013 and CK data sets to produce cartoon expression data sets, and trains and compares cartoon data sets and original data sets respectively under the same network. The experimental results show that the method proposed in this paper has high detection accuracy and fast speed on standardized and neat data sets. On the data set with more unfavorable factors, the training accuracy of the two methods is similar, but the processing speed of the proposed method is faster. Experimental results show that the proposed method is feasible and effective.