International Journal of Information Technology最新文献

The Internet of Things (IoT) has recently become an important innovation in building smart environments. With any technology that relies on the Internet of Things model, security and privacy are seen as key issues. Many privacy and security concerns arise due to the various possibilities of intruders to attack the system. Due to the dynamic and heterogeneous nature of IoT devices and networks, we propose a novel approach for attack detection in IoT environments by combining two modifications based on the Adaptive Neuro-Fuzzy Inference System (ANFIS). For the efficient extraction of features from input datasets, we use a Relational Auto Encoder (RAE) Network, followed by an enhanced version of the ANFIS model. ANFIS parameters have been optimized to use Gaussian kernel membership functions and the Enhanced Osprey optimization algorithm (EOOA) has been used to optimize initial ANFIS parameters. As part of the experimental analysis, two sets of datasets are used; these are NSL-KDD 99 and UNSW-NB15 datasets, which contain different kinds of attack labels such as DoS, probing, U2R, and R2L attacks. Performance metrics including accuracy, precision, recall, and F-measure are used to assess the effectiveness of our proposed scheme. As a result of this approach, we have demonstrated promising results in identifying attackers for IoT security applications, while also offering robustness and scalability.

As ICT technology has developed, work has become possible in a variety of locations and working from home has become more active. Intranet-type information network access was physically connected within the corporate building. Currently, access to the Internet is possible from outside, regardless of geographical location. Because of this, in addition to strengthening internal security, numerous studies are being conducted on external threat factors, user authentication, and data security. However, sophisticated attacks require security technologies such as enhanced network access control and strict user authentication. In this study, we propose an Abnormal Behavior Detection Mechanism (ABDM) that analyzes packets for various purposes for external access and determines abnormal behavior using a zero-trust perspective. ABDM approached users, systems, and time series to analyze packets and determine abnormal behavior. As a result, an accuracy of approximately 93% for abnormal behavior was measured.

This paper presents a ternary D-latch design using resistive random-access memory (RRAM) and carbon nanotube field effect transistor (CNTFET) technology. The property of multi-threshold in CNTFETs and multi-level cell in RRAM is utilized in designing ternary logic circuits. The advantages of ternary logic provide best substitute to replace conventional binary logic system such as less interconnect complexity, enhanced information density, compact chip area and fast computational ability. As a result, the ternary system offers digital designs that are easy to implement while maintaining both high energy efficiency and rapid signal processing. This paper presents a ternary D-latch circuit utilizing CNTFET-RRAM based ternary logic gates including standard ternary inverter (STI) and ternary NAND (TNAND). The proposed design provides 0.863 nW power consumption and 12 ps delay.

Cancer has a high mortality rate due to its aggressiveness, potential for metastasis, and heterogeneity. There are several types of cancers that are common throughout the world, including lung cancer and colon cancer. Radiologists have developed several expert systems to assist in the diagnosis of lung cancer over the years. However, this requires accurate research. Therefore, in this paper, an automatic lung and colon cancer classification model based on a machine learning algorithm is proposed. Initially, the histopathological images are collected from the dataset. Then, to reduce the noise present in the input image, we apply the adaptive median filter. After noise removal, we use novel feature extraction techniques, gray-level histogram (MGH) of moments, local binary pattern (LBP) features, and gray-level co-occurrence matrix (GLCM) features and morphological features to extract features. Since the large number of features is a major obstacle to the classification process, improved local Fisher discriminant analysis (ILFDA) is used to reduce the dimensionality after feature extraction. After feature selection, the selected features are given to an enhanced ANFIS classifier to classify an image as normal or abnormal. The performance of the proposed approach is analyzed based on different metrics. The proposed method is implemented in Python.

Individuals with food allergies face limitations in social events and restaurant dining. Artificial intelligence solutions should be offered to this category. In this paper, a recommender system is proposed for the benefit of people with food allergies. The system aims to identify convenient options for the user in a restaurant/hotel menu. The system collects user’s allergy information and the restaurant menu, it extracts dishes names using a machine learning model. Then it conducts search about the recipes of these dishes and identify allergen-free ones. The system has been implemented as a mobile application involving a Naïve Bayes classification model and a web search API. The performance of the classifier was significant (accuracy 87%). Yet, an enhancement approach was introduced to increase the accuracy to 90%. In addition, an expert-driven test has been conducted and 98.5% of the system allergen identification was accurate in comparison with the original recipes used by restaurants’ chefs.

Misinterpreting satirical posts can contribute to the spread of misinformation and potentially be a source of what is commonly referred to as “fake news”. Satire is a form of humor that often involves exaggeration, irony, or ridicule to comment on or criticize a particular subject. While satirical content is not intended to be taken literally, there are instances where individuals may misinterpret it, leading to the dissemination of false information. In fact, we can reduce the spread of fake news by preventing people from misinterpreting satirical posts. However, sarcasm recognition is considered a challenging task in the Sentiment Analysis domain. Even for humans, it can be difficult to recognize irony and sarcasm, which conveys a sharp, bitter remark or criticism in ambiguous and unclear natural language. This makes the identification much more difficult for an automated model. In this paper, we have carried out an in-depth literature review about the main approaches used for sarcasm detection and especially those based on Machine Learning (ML) models. Then, a study was conducted with a series of binary classification models that exploit a variety of statistical and semantic features. Our experiments have been carried out on twitter dataset obtained from SemEval-2018 Task 3. An extensive evaluation of each set of classifiers demonstrates the efficiency of our proposed model in detecting and identifying sarcastic content in tweets. Finally, we compared the performance of machine learning models using our proposed features with our baseline and state-of-the-art on the same dataset. By using Support Vector Machine (SVM) model and the proposed features, we outperform the state-of-the-art and we obtained an accuracy of 79.46% with a F-score equal to 79.66% which considered a promising result in this field.

Mental health disorders present a significant global challenge, requiring early detection for effective intervention. This research explores the comparative performance of two advanced hybrid classifiers against conventional machine learning models. Introducing an innovative hybrid classifier framework, we combine Support Vector Machines with Neural Networks (Hybrid_1) and Random Forests with Gradient Boosting Machines (Hybrid_2), creating synergistic combinations of traditional and ensemble learning approaches. Using this novel fusion technique, we conduct a comprehensive analysis, emphasizing customized feature engineering strategies tailored for mental health assessment. Evaluation on the Mental_health dataset demonstrates the superior performance of hybrid classifiers, achieving accuracy rates of 86.69% and 93.54% for Hybrid_1 and Hybrid_2, respectively. These results highlight the potential of hybrid classifiers in mental health prediction and emphasize the crucial role of feature engineering in model optimization. Our pioneering hybrids, Hybrid_1 and Hybrid_2, represent a breakthrough, seamlessly integrating Support Vector Machines with Neural Networks and Random Forests with Gradient Boosting Machines, respectively. Distinguished from conventional approaches, our hybrids leverage the combined strengths of diverse algorithms, addressing challenges associated with complex feature relationships and dataset adaptability. This study not only showcases the promise of hybrid classifiers in mental health assessment but also provides valuable insights into feature selection and model interpretability, enhancing our understanding of this critical domain.

Supply chain management involves managing the entire manufacturing process, from purchasing supplies to delivering the final product. Demand forecasting helps businesses predict future customer demand by analyzing historical data and market patterns. While various papers discuss optimizing models, this research compares several machine learning models, such as ARIMA, SARIMA, and deep learning models like RNN, LSTM, GRU, and BLSTM. It also extends to approaches like ensemble learning with the LSTM model, discussing how ensemble learning can further improve the LSTM model. This paper explores ensemble learning in two ways: a) without model pruning, averaging all generated models, and b) with model pruning, removing underperforming models and averaging top performers. Experiments conducted on a public dataset from the University of Chicago achieved a very low RMSE loss of 9.26 on the LSTM model improved via ensemble learning with model pruning. This ensemble approach with model pruning improved accuracy in predicting future customer demand, and a complete pipeline integrating visualization and a notification system was developed.

WDM system transmission efficiency is deteriorated by the combined influence of cross phase modulation (XPM) and group velocity dispersion (GVD) of first and second order. This degradation occurs as the channel bulk, light intensity, speed of transmission, and wavelength count frequencies increase. Analysis of the pulse broadening factor, standardized outturn, and resolving the nonlinear Schrödinger equation (NLSE) is conducted in this study. The influence of XPM on higher order GVD is reflected. The impact of broadcast limit and different absorbed powers (10 mW to 120 mW) at various transmission speeds (10 Gbps and 40 Gbps) are assessed utilizing large effective area fiber (LEAF) and standard single mode fiber (SSMF). The first- and second order GVD XPM impacts are the only ones that influence emitted oscillation. GVD's second-order consequences are not perceptible at close grips (⁓10 km) and low throughput (10 Gbps) but become perceptible and affect system performance at bit rates of 40 Gbps and above. The study found that transmission rate and fiber span have a stronger impression on duration than input dominance. The SSMF and LEAF consequences are obtained by rigorous derivation and numerical simulation at the consistent throughput and absorbed power managing the split-phase Fourier method. XPM has a stronger optimistic impact on GVD in SSMF fibers than LEAF fibers by 2 km. Due to their ability to quantify the degree of performance degradation emanating from XPM effects with first- and second order GVD, the findings of this work may prove useful in the design of high-speed, long-distance WDM fiber-optic transmission links.

Epilepsy is a chronic neurological disorder characterized by the occurrence of paroxysmal recurrent seizures, which are caused by abnormal electrical activity in the brain. Seizures vary widely in their presentation, depending on the specific region of the brain involved and the extent of the abnormal electrical discharges. The disease can affect cognitive function posing a serious threat to the patients’ lives. Epilepsy causes emotional and behavioral changes, along with sleep disorders and migraines, leading to social isolation and discrimination. Timely administration of medication can cure most cases of epilepsy. However, identifying epileptic patients requires reviewing multiple EEG signal sheets, which can delay disease prediction. Therefore, the aim of our study is to apply simplistic machine learning algorithms that can study the EEG signal data swiftly to identify individuals in seizure, inter-ictal, and normal states, thereby aiding in medical diagnosis. The novelty of this study lies in the utilization of pre-built methods and develop a fast and efficient model that is lightweight and easy to integrate in healthcare to provide relief to epileptic patients. While previous studies have achieved high accuracy, the discussion involving time complexity of their models has been scarce. Given the importance of timely medication in managing epilepsy, it is crucial to consider the runtime of the model rather than solely focusing on accuracy. Therefore, a model that balances both a short runtime (2.9 min) and a satisfactory accuracy (97.46%) has been developed in this project. Integration of this project's findings will catalyze transformative changes within the healthcare industry, enabling healthcare professionals to detect epilepsy at earlier stages and provide timely interventions, ultimately fostering a system that prioritizes precision, innovation, and improved patient outcomes.