ABSTRACT: 

Information Security experts have been focusing on the study of malwares detection because of its rise recently. Android is the overwhelmingly dominant mobile operating system on the African phone market with 85.88%, while  iOS, Series 40, Windows and other Operating systems has 8.84%, 0.57%, 0.39% and 3.23% respectively. Cybercriminals uses various types of malware to attack and compromise android devices in the recent years with the intent of getting access to sensitive information present in the victim’s devices. Various detection tools exits in the market, but how reliable in terms of performance were put to experimental analysis in this work. Five antimalware tools and eight malwares were used for our analysis under a virtual environment. Results of our analysis shows kaspersky antimalware as best performed while 360 Total Security as least performed as it failed to detect any of the malware samples despite its high rating performance in apps store.

Keywords: Cybercriminals, Android, Antimalware, Malware, Mobile device

ABSTRACT: 

Recent studies have brought to light a concerning trend: a threefold increase in smishing attacks in comparison to previous years. These malicious activities present a significant threat to individuals who often remain oblivious to the associated risks. The manipulative nature of smishing makes it challenging for recipients to distinguish between legitimate SMS messages and phishing attempts. To address this growing issue, various research endeavors have focused on devising effective detection techniques, resulting in a range of accuracies and false positive rates. This paper proposes a stacked ensemble machine learning approach designed to tackle smishing attacks by analyzing the features contained within The SMS Smish Collection dataset v.1. Recognizing the importance of balanced data for training robust models, this paper employs the Synthetic Minority Oversampling Technique (SMOTE) to rectify class imbalances and the primary detection model is a stacked ensemble learning model, which combines the strengths of two different algorithms: Random Forest, and Gradient Boosting. To evaluate the effectiveness of our approach, we conduct a rigorous 70:30 data split. The results of our experiments demonstrate an accuracy (ACC) of 99.83%, coupled with an impressively low false positive rate (FPR) of 0.039. Beyond these metrics, we also measure other vital criteria such as precision, recall, and F1-measure to provide a comprehensive assessment of the stacked ensemble model's performance. In conclusion, our research addresses the critical issue of smishing attacks, which are on the rise and pose a significant threat to individuals' cybersecurity. By leveraging machine learning techniques and a carefully curated dataset, our approach achieves an outstanding level of accuracy in detecting smishing attempts. We believe that our findings contribute significantly to the field of cybersecurity and have practical implications for safeguarding individuals from the growing menace of smishing attacks. Furthermore, this paper underscores the significance of continuous monitoring and refinement of the stacked ensemble machine learning approach, as cybercriminals are constantly devising new tactics and techniques, necessitating ongoing advancements in detection methodologies to stay ahead of emerging attacks and threat.

Keywords: Machine Learning; Smishing Detection; Stacked; Ensemble Machine Learning; SMS

ABSTRACT: 

In the recent LLM boom, a collaboration of multiple modalities such as sentence and images are attracting attention. To develop an original system, it is necessary to understand the basics of learning and inference methods. Therefore, a basic multimodal system was constructed and investigated changes in prediction accuracy when changing the learning model. Such basic surveys will form the high-performance basis technology for building the metaverse application.

Keywords: Multimodal learning, predict accuracy, learning model, Deep learning.

ABSTRACT: 

Digital marketing is critical to any company's marketing strategy, regardless of sector, size, or country of origin. As a

result, companies are being forced to use this type of marketing more than ever before in order to remain competitive. It essentially can bring huge benefits at low costs. The most common type of digital marketing is inbound marketing, which is an organic marketing form based on a close relationship between the company and its prospects or customers who have voluntarily expressed their interest in the company's products (via subscription to newsletters, blogs, social networks, etc.) and who have been attracted and involved by high-quality content marketing.

Keywords: Digital marketing, SEO, Inbound marketing, Email marketing, Social networks.

ABSTRACT: 

The stream cipher is an important branch of modern cryptography that is used in a variety of applications and devices such as the A5 algorithms for GSM, E0 for Bluetooth, Secure socket layer(SSL) for WEB, and so on. The driving part of the stream cipher systems is a set of Linear Feedback Shift Registers (LFSR). The output of these (LFSR) compounds by the combiner function G is used to generate the Running key generator (RKG) keysequence. The recurrence relation is used to determine the next state values of the LFSR based on initial values. As a result, the construction of a linear equations system of the LFSR can be considered, and RKG attack means attempting to find the initial values of the driving part's LFSRs. In this paper, a three-stage approach was used, beginning with the construction of a linear equations system of an LFSR, followed by testing the uniqueness of the solution, and finally solving the linear equations system using one of the classical methods Gauss elimination, the solution of which is the initial values of the generator's LFSRs. One well-known algorithm, the Geffe generator, was used as a case study.

Keyword: stream cipher, RKG, Feedback function, Linear Equation System, LFSR, driving part, Combiner function G