Doctoral Dissertations
https://hdl.handle.net/10155/402
Doctoral Dissertations (FSCI)2024-03-29T00:58:17ZDL-based defense against polymorphic network attacks
https://hdl.handle.net/10155/1759
DL-based defense against polymorphic network attacks
Sabeel, Ulya
Network security is of vital importance in our world dominated by internet systems. These systems are vulnerable to large-scale rapidly evolving attacks by sophisticated cyber attackers who can have an upper edge over the defensive systems. Artificial Intelligence (AI) based intrusion detection systems provide effective defense mechanisms against cyber attacks. However, these techniques often rely on the same dataset for training and validation as well as evaluation of AI models. Current research [1] also confirms that such trained models can accurately identify known/typical network attacks but perform poorly when faced with continuously evolving atypical/polymorphic cyberattacks. Therefore, it is crucial to develop and train an AI-based Intrusion Detection System (IDS) that proactively learns to resist infiltration by such dynamically changing attacks.
For this purpose, in this research work, we propose an AI-based IDS system that can monitor and detect polymorphic network attacks with high confidence levels. We propose a novel hybrid adversarial model that leverages the best characteristics of a Conditional Variational Autoencoder (CVAE) and a Generative Adversarial Network (GAN). Our system generates adversarial polymorphic attacks against the IDS to examine its performance and incrementally retrains it to strengthen its detection of new attacks, specifically for minority attack samples in the input data. The employed attack quality analysis ensures that the adversarial atypical/polymorphic attacks generated through our system resemble realistic network attacks. Our experiments showcase the exceptional performance of the proposed IDS by training it using the CICIDS2017 and CICIoT2023 benchmark datasets and evaluating its performance against several atypical/polymorphic attack flows. The results indicate that the proposed technique, through adaptive training, learns the pattern of dynamically changing atypical/polymorphic attacks and identifies such attacks with high IDS proficiency. Additionally, our IDS surpasses various state-of-the-art anomaly detection and class balancing techniques.
2024-01-01T00:00:00ZIdentification and characterization novel of cystine-loop ligand- gated chloride channels from Dirofilaria immitis: pharmacological analysis and novel compound screening
https://hdl.handle.net/10155/1755
Identification and characterization novel of cystine-loop ligand- gated chloride channels from Dirofilaria immitis: pharmacological analysis and novel compound screening
Varley, Sierra
Dirofilaria immitis, otherwise known as heartworm, is a parasite that infects the hearts of dogs and causes serious health consequences. While preventative treatments are available drug resistance is developing at an alarming rate. Cystine-loop ligand-gated ion channels are important receptors in nematode neurobiology, and as such are promising drug targets. The UNC-49 (GABA-gated chloride channel) and ACC (acetylcholine-gated chloride channel) family of receptors have been characterized as potential drug targets in other nematodes. However, these receptors have yet to be identified or characterized in D. immitis. This thesis investigates the cloning and pharmacological characterization of 5 novel receptor subunits from D. immitis: UNC-49B and UNC-49C, ACC-1, LGC-46 and LGC-47. Additionally, novel derivatives of the antiparasitic drug levamisole were tested on ACC receptors to determine if any modifications enhanced levamisole action. D. immitis UNC-49B assembled as a homomeric channel and exhibited an EC50 of 5mM for GABA. UNC-49B also formed a functional channel with UNC-49C, which exhibited a decrease in GABA sensitivity. Additionally, the D. immitis UNC-49 receptors were significantly more sensitive to the open channel blocker picrotoxin compared to the same receptors from the sheep parasite Haemonchus contortus. Moreover, D. immitis UNC-49C, unlike other UNC-49C subunits, did not cause a decrease in picrotoxin sensitivity when assembled with UNC-49B. D. immitis ACC-1, LGC-46, and LGC-47 were unable to form functional channels through heterologous expression in Xenopus laevis oocytes. Levamisole derivatives were therefore tested on H. contortus ACC-2, which identified at least one showing a higher sensitivity compared to levamisole. Overall, this study characterized 2 novel UNC-49 receptors, identified 3 potential members of the ACC family in D. immitis and tested 8 novel derivatives of levamisole. This study lays foundation for the identification of more ligand-gated ion channels and will serve as a starting point for future researchers looking at new drug potential targets in D. immitis.
2023-12-01T00:00:00ZLeveraging vehicular cloud computing through location and request prediction
https://hdl.handle.net/10155/1614
Leveraging vehicular cloud computing through location and request prediction
Miri, Farimasadat
In recent years, the implementation of Vehicular Ad-hoc networks (VANET) has been acknowledged as a promising solution for monitoring road conditions. Despite the many benefits they can bring to society, the growing demand for communication, storage, and processing capabilities is giving rise to new challenges. For instance, the heightened need for communication in VANETs can cause network congestion. Additionally, the real-time nature of VANET applications, such as traffic management, accident prevention, and navigation, necessitates rapid and reliable communication, which can be difficult to achieve in a network that is constantly changing. Moreover, managing and scaling the resources of a large scale network like VANET with a large number of vehicles and road-side units (RSUs) is a significant challenge. There are essential contributions to deal with these problems, such as utilizing MEC (Mobile Edge Computing), and a hybrid architecture of cloud and fog computing which can create an efficient and adaptable resource management system. However, having unpredictable events such as accidents or road closures can cause rapid changes in the network topology, making it difficult to allocate resources effectively. Also, unpredictable events can lead to a lack of information, making it difficult to obtain accurate and up-to-date information about the network and hard to allocate resources effectively. Furthermore, allocating resources at the right time when unpredictable events happen without network congestion is another challenging problem that causes us to think about proposing a model that can satisfy delay sensitive applications requirements and decrease the monetary cost in hybrid cloud and fog architecture in the presence of unpredictable congestion. To reach our goal, we first need to estimate the traffic flow after an accident, predict the level of congestion based on the number of requests from different vehicles, and then predict the location of potential vehicles as mobile fog nodes in advance to form Vehicular Clouds. Finally, we propose a layer-based architecture with two prediction models and different modules to support safety and non-safety applications, and a task scheduling mechanism to decrease monetary costs and delay for delay-sensitive applications during congestion times to serve the vehicles requests in that region.
2023-04-01T00:00:00ZIntegrated traffic analysis and visualization for future road events
https://hdl.handle.net/10155/1594
Integrated traffic analysis and visualization for future road events
Alghamdi, Taghreed
The existing traffic simulation methods are limited to specific synthetic scenarios. In addition, the natural structure of traffic and accident data requires modeling the dependent observations on multiple levels. Therefore, a system that utilizes hierarchical LMMs and GBM models are proposed which adaptively analyzes and predicts the traffic pattern based on hypothetical inputs. We developed a user-friendly interface to show the outcomes of the hybrid model. The proposed system encompasses three major components: (1) a road accident simulator and event profile to simulate an accident and predict its effects on traffic status; (2) a robust spatiotemporal traffic speed prediction model that integrates the impact of road accident with the prediction model to adaptively predict the future traffic status in response to this accident; (3) a traffic simulation tool to present the future traffic status. Our system provides satisfactory prediction results in terms of predicting with small errors, obtaining optimal hyperparameters, and less computational complexity.
The hierarchical structure of the spatial component in our approach effectively captures the correlation in traffic status across different spatial points on the same road. Furthermore, computing the traffic speed at different spatial levels and how it interacts with lagged prior traffic speed over the past four periods and a day prior up-scaled the system efficiency.
Evaluation is conducted to test the functionality, usability, and viability. Performance evaluation shows that the event profile model achieves small error rates with an MSE of 0.24 and an RMSE of 0.53 on the testing data, demonstrating satisfactory performance. For traffic status, the integrated model achieves high accuracy with low computational complexity. The boosted LMMs achieved high performance on the test data with an R2 of 0.9190 and an R2 of 0.9291 on the full-fitted dataset. The MAE and RMSE are 0.27 and 0.80, respectively, indicating that the fitness of our data was excellent.
2023-04-01T00:00:00Z