http://hdl.handle.net/123456789/13167 2026-04-04T10:53:34Z http://hdl.handle.net/123456789/19737 Exploring Evolutionary Algorithms for Optimal Features Selection to Detect Anomaly Based Intrusion In IoT Ali Arshad, 01-243231-002 In light of the fast-evolving scenario of the IoT, network security constantly bears immense challenges due to the increasing number of cyber threats and vulnerabilities. Traditional intrusion detection systems use predefined signatures and rule-based approaches to detect malicious activities within a network. In contrast to the traditional signature-based intrusion detection system, which utilizes previously established attack patterns, an anomaly-based intrusion detection system typically utilizes machine learning, statistical models, and artificial intelligence to assess network traffic, system log entries, and user behavior. . This research proposes the anomaly-based intrusion detection system using Particle Swarm Optimization (PSO)-based feature selection and ensemble learning in enhancing detection accuracy for IoT networks. Performance of stacking, hard voting, soft voting, and autoencoder-based models is evaluated over the benchmark datasets NSL-KDD and KDDCup 99 in analyzing their effectiveness in detecting anomalous behaviors in IoT environments. From the results, it is clear that PSO-based feature selection is highly significant in anomaly detection. Anomaly detection gets better performance by reducing feature redundancy along with improving classification accuracy. Of all the models tested, Stacking performed the best, with an accuracy of 98.87% on NSL-KDD and 99.76% on KDDCup 99, proving to be the most effective method. Soft Voting and Hard Voting also did well on NSL-KDD, recording 98.38% and 98.13% accuracy respectively, which highlighted the strength of ensemble methods in identifying anomalies based on IoT. The Autoencoder model demonstrated unsupervised anomaly detection, with a slightly less accurate of 96.63% on NSL-KDD and 98.23% on KDDCup 99, due to its greater false positive rates. This suggests that models based on deep learning could make a significant performance improvement on anomaly classification by leveraging their explicit feature selection capabilities. Stacking is an ensemble learning algorithm that increases predictive performance by aggregating multiple base models in a superior way. It can be used optimally when there exists some labeled data in supervised learning cases. Auto-Encoders correspondingly are neural networks deployed mainly in unsupervised learning, and they serve very well in detecting anomalies without labeled data. Supervised by Dr. Saba Mahmood 2025-01-01T00:00:00Z http://hdl.handle.net/123456789/19598 Optimized Software Implementations for NIST’s Post Quantum Cryptography Candidates Using NEON Based Instructions Anjum Ashraaf, 01-247231-005 The development of quantum computers is a serious threat to traditional cryptography methods that use well-known algorithms like RSA and ECC. These algorithms are secure from classical computers, but they rely on mathematical problems like integer factorization and discrete logarithm problems, which quantum computers can solve effectively. To counter this new threat, cryptographers throughout the world are actively developing a set of quantum-resistant cryptographic protocols called Post-Quantum Cryptography (PQC). The computational effectiveness and performance of PQC algorithms on low-resource devices, however, pose a serious barrier to their widespread use. This research aims to optimize post-quantum cryptographic algorithms, specifically NTRU and Kyber, for ARM-based systems. The study focuses on implementing these algorithms on Android (Google Pixel 2) and Raspberry Pi while leveraging NEON SIMD optimizations to enhance efficiency. A key focus is optimizing polynomial multiplication, a core operation in lattice-based encryption, by evaluating various strategies and selecting the most effective for NEON acceleration. Performance improvements are systematically analyzed across different versions of Kyber and NTRU, with benchmarking against NIST PQC reference implementations to assess speedup and efficiency gains. This research presents the first NEON-optimized implementations of post-quantum cryptographic algorithms on Android, achieving significant performance gain of 2.6x in encryption and 3.1x in decryption of kyber on Android. Similarly in case of NTRU the performance gain is 2.6x in encryption and 5.5x in decryption when compared with the reference implementation. Additionally, a new multiplication design is also presented for kyber on Intel processors, which gives performance gain of 1.17x in encryption and 1.19x in decrypiton suggesting similar optimization techniques could enhance PQC performance on ARM architectures. A novel variant of the NTRU cryptosystem, called M-NTRU, is introduced. NTRU relies on polynomial arithmetic over Z[x] and the hardness of the closest and shortest vector problems, with security based on solving polynomial equations modulo unrelated moduli. In M-NTRU, the polynomial ring Z[x] is replaced with the matrix ring Zp[x]/⟨Xn+1⟩, operating in a non-commutative matrix ring within a Galois field. This enhances security by expanding the key space and mitigating lattice attacks due to highdimensional structures and non-commutativity. While maintaining the conventional NTRU framework, M-NTRU modifies polynomial representation, arithmetic operations, and key generation to strengthen resistance against cryptanalysis. These optimizations will benefit greatly in adoption of the PQC algorithms specially on the resource constrained ARM devices. Supervised by Dr. Asim Ali 2025-01-01T00:00:00Z http://hdl.handle.net/123456789/19597 A Hybrid Authentication Approach (LPCAP-IoD) To Enhance Security in Internet of Drones (IoD) Muhammad Imran Khan, 01-247231-008 The Internet of Drones (IoD) is a field where a lot of research is underway due to the ever growing demand and applications in all domain of life such as miliary (for Information sharing, search & rescue, surveillance, imagery, payload transfer etc), socio-economic development, agriculture, shipment/ delivery services, disaster management and fire fighting. However, the open and dynamic nature of IoD environments makes them vulnerable to various security attacks/ threats such as replay attacks, impersonation attacks, eavesdropping, drone hijacking, physical attacks, quantum computing threats etc. Various schemes have been proposed by the researcher to enhance security aspects related to authentication protocol of IoD. However, all have their short coming due to a trade off between security features and computation/ communication costs. This thesis proposes a hybrid authentication protocol for IoD that combines Lattice-based Cryptography, Chebyshev Chaotic Maps and Physical Unclonable Function (PUF) and provide resistance against advanced security attacks/ threats including quantum computing threats. The proposed scheme has been termed as LPCAP-IoD (Lattice, PUF and Chaotic Maps based Authentication Protocol - Internet of Drones). The formal and informal comprehensive security testing of proposed protocol has been undertaken by utilizing tools like ProVerif an NS-3. Furthermore, efficiency of the protocol has been determined by calculating the desired computational and communication costs and comparing them with the existing authentication schemes. The results of our formal and informal security testing demonstrate that the protocol is resistant to Quantum Computing Attacks as well as against a wide range of advance attacks and is suitable for IoD environment where security is paramount. Future research directions are also discussed to further enhance the security and scalability of our proposed protocol. Supervised by Dr. Muhammad Khurram Ehsan 2025-01-01T00:00:00Z http://hdl.handle.net/123456789/19599 Network Attacks Classification and Prediction: Exploring Advanced Strategies for Enhanced Network Security Furqan Aziz Qureshi, 01-247231-006 In the age of increasing cyber attacks, strong network security is an ongoing challenge. This research uses the UNSW-NB15 dataset to create and test advanced machine learning methods for the efficient classification and prediction of network attacks. The study begins with careful data preprocessing, which involves outlier detection and mitigation, as well as the use of Borderline SMOTE to solve class imbalance. With a primary focus on binary classification, the study performs comparative analysis of some ensemble techniques. The Random Forest classifier was found to be the best performer with an accuracy of 93% and an F1-score of 0.93, thus surpassing past benchmarks where its accuracy was stated at 90%. Furthermore, multi-class classification experiments were conducted to further confirm the method and prove its usability in different attack scenarios. These results highlight the power of sophisticated classification methods in strengthening network defense mechanisms and offer significant insights for future studies in the field of cyberattack prediction and prevention. Supervised by Dr. Moazam Ali 2025-01-01T00:00:00Z