The rapid advancement of deepfake technology presents significant challenges in detecting highly convincing fake videos, posing risks such as misinformation, identity theft, and privacy violations. In response, this paper proposes an innovative approach to deepfake video detection by integrating features derived from ant colony optimization–particle swarm optimization (ACO-PSO) and deep learning techniques. The proposed methodology leverages ACO-PSO features and deep learning models to enhance detection accuracy and robustness. Features from ACO-PSO are extracted from the spatial and temporal characteristics of video frames, capturing subtle patterns indicative of deepfake manipulation. These features are then used to train a deep learning classifier to automatically distinguish between authentic and deepfake videos. Extensive experiments using comparative datasets demonstrate the superiority of the proposed method in terms of detection accuracy, robustness to manipulation techniques, and generalization to unseen data. The computational efficiency of the approach is also analyzed, highlighting its practical feasibility for real-time applications. The findings revealed that the proposed method achieved an accuracy of 98.91% and an F1 score of 99.12%, indicating remarkable success in deepfake detection. The integration of ACO-PSO features and deep learning enables comprehensive analysis, bolstering precision and resilience in detecting deepfake content. This approach addresses the challenges involved in facial forgery detection and contributes to safeguarding digital media integrity amid misinformation and manipulation.

Chronic Otitis Media (COM) causes deformation of the middle ear ossicles with perforation as a result of long-lasting inflammation of the middle ear and it is one of the basic reasons for hearing loss. The middle ear images are examined by otolaryngologists in the diagnosis of the disease in clinical practice. The observers make a decision considering the status of the tympanic membrane images. Decision support systems using image processing techniques and machine learning algorithms are quite useful in the diagnosis process, however, the usage of such systems in this field is limited. In this study, we propose a diagnostic model using a pretrained deep convolutional neural network (DCNN) called AlexNet. The experiments were carried out on a private dataset consisting of totally 598 tympanic membrane images collected from patients admitted to Özel Van Akdamar Hospital. Firstly, a set of preprocessing procedures were applied to the eardrum images. Then, the tympanic membrane images were used to feed the DCNN model. The proposed model was trained using transfer learning approach. To evaluate and validate the success of the proposed model, the 10-fold cross-validation method was used. As a result, the model provided satisfactory results with an accuracy of 98.77%. Consequently, the proposed DCNN model was determined as a robust tool in separating chronic and normal tympanic membrane images.

Intrusion detection systems utilize the analysis of log data to effectively detect anomalies. However, detecting anomalies quickly and effectively in large and heterogeneous log data can be challenging. To address this difficulty, this study proposes the GLSTM (Graph-based Long Short-Term Memory) framework, a graph-based deep learning model that analyzes log data to detect cyber-attacks rapidly and effectively. The framework involves standardizing the complex and diverse log data, training this data on an artificial intelligence model, and detecting anomalies. Initially, the complex and diverse log data is transformed into graph data using Node2Vec, enabling efficient and rapid analysis on the artificial intelligence model. Subsequently, these graph data are trained using LSTM (Long Short-Term Memory), Bi-LSTM, and GRU(Gated Recurrent Unit) deep learning algorithms. The proposed framework is tested using Hadoop’s HDFS dataset, collected from different systems and heterogeneous sources, as well as the BGL and IMDB datasets. Experimental results on the selected datasets demonstrate high levels of success.