http://hdl.handle.net/123456789/21324 Sun, 20 Sep 2026 22:52:34 GMT 2026-09-20T22:52:34Z http://hdl.handle.net/123456789/21644 Cyber Sentinel: AI-Powered Anomalous Behavior Detection System Muhammad Usman Iftikhar, 01-136221-022; Muhammad Hermen Khan, 01-136221-054 User authentication has been and still remains the most exploited of all weak points in the digital worlds of today. Cybersecurity incidents. The project named "CyberSentinel: AI-Powered Anomalous Behavioral Detection System" introduces a smart authentication method that consists of three factors working together to make system access more secure by the simultaneous application of password checking, face recognition, and keystroke dynamics reporting. CyberSentinel, as opposed to traditional login systems that either depend entirely on passwords or on the use of a single biometric trait, integrates the use of artificial intelligence in a way that it has three levels of verification. The first level checks the user’s identity by comparing the password. The second level applies a face-examination model that obtains a still image at the time of login to check the rightful user through deepface and spoof detection to block facial impersonation. The third level depends on biometrics of behavior by exploring users’ keystroke patterns like typing speed, holding time, and latency with a Long Short-Term Memory (LSTM) neural network. The system was developed in Python and made possible by the use of the most advanced frameworks including TensorFlow, OpenCV, and Scikit-learn. The backend is supported by FastAPI (Fast Application Programming Interface) and is connected to a ReactJS (React JavaScript Library) based user interface. The system will use a fusion-based decision mechanism which means it will check the results coming from all three modules and allow access only when there is an agreement among them. This proposed structure is good enough to detect strange login attempts, impersonations, and wrongful access, and still be user-friendly. Merging physiological and deepface for emotion detection along with AI-powered analytics, CyberSentinel is a secure, adaptive, and efficient authentication solution that lessens the impact of the inherent drawbacks of single-factor systems. Supervised by Dr. Sadia Nazim Wed, 01 Jan 2025 00:00:00 GMT http://hdl.handle.net/123456789/21644 2025-01-01T00:00:00Z http://hdl.handle.net/123456789/21645 Fintrix: An Agentic AI Powered Customer Service Solution for the Banking Sector Saim Chishti, 01-136221-045; Muhammad Taha Hasnat, 01-136221-018 The fast pace of adoption for AI is creating a new landscape in financial services where intel ligent automation, improved decision-making, and powerful defenses against increasingly sophisticated digital attacks are now possible. This thesis introduces Fintrix, anAgentic AI-wielding intelligent banking assistant that consolidates various AI functionalities in a unified, self-contained entity. Fintrix is composed of three main modules: (i) a Customer Support Agent which provides efficient, contextual and personalized counseling using advanced natural language understanding; (ii) a Financial Advisory Agent, based on predic tive analytics, time series modeling and behavior analysis that offerspersonalized financial advice and insights; nd (iii) a Fraud De- tection Agent which employs machine-learning techniques together with anomaly-detection methods to recognize fraudulent activities in real-time. Fintrix is grounded on modular, scalable and explainable (XAI) methods for natural language understanding with the intuition behind Agentic AI, Large Language Models (LLMs), andXAI approaches. These are just a few possible advantages of suchan inter-relation of operations, which provides the systems with self-management to per form ultra complex economic operations on their own responsibility and anyway always give content clear and understandable to the end user. Synchronous running of multiple agents "Fintrix ensures that rendezvous between the state rooms can be accomplished bad decision making and overall accounting overhead onthe bank’s computer. In summary, Fintrix demonstrates the practicalitiesand relevance of a decentralised MAS approach in the modern financial sector. It also demonstrates how Agentic AI drives improved customer service, better riskmanagement and data-driven financial planning Supervised by Dr. Faryal Nosheen Wed, 01 Jan 2025 00:00:00 GMT http://hdl.handle.net/123456789/21645 2025-01-01T00:00:00Z http://hdl.handle.net/123456789/21646 Neuro Motion Wheelchair Guidance Eman Ali, 01-136221-006; Akasha Hashmi, 01-136221-003 Neuro Motion Wheelchair Guidance is a smart self contained assistive mobility assistance system that enables the process of empowering persons with severe physical disabilities by incorporating brain computer interface (BCI) and gesture recognition systems. The system allows users to operate a motorized wheelchair by using the brainwave too as well as the predetermined gestures with the help of the hands without any additional equipment or being connected to the internet. The suggested solution will make use of a Raspberry Pi 5 as the core unit of the processing machine, to which an EEG headset (to collect neural signals) and a camera module (to record hand gestures) will be connected. Two special AI models, which are trained on the EEG and visual gesture datasets, are run locally on the Raspberry Pi to categorize user commands in a real-time. According to these forecasts, the L298 motor controller will act as the motor controller that drives the forward and backward movement of the DC motors with a 12V supply and a 5V powered servo motor will be used to steer to the left and right. Every element motors, sensors, and controllers are also integrated together using a smooth interface using the GPIO and are also minimized to achieve low latency response. It is a fully autonomous hands free mobility system that is designed to provide an experience of complete autonomy in mobility compared to traditional joystick or voice based control systems, which are recommended to patients with such conditions as ALS, cerebral palsy, or spinal cord injuries. The built-in operating system also avoids any reliance on the mobile applications or network connectivity and guarantees regular operation and privacy of the user. The Neuro Motion Wheelchair creates a solid and compatible platform of the future generation of assistive robotics by combining neural computing with embedded AI. It shows how the availability of technology and intelligent automation can help people with physical impairments be independent, have a high level of mobility, and have the quality of life that is highly improved. Supervised by Mr. Abdul Rahman Wed, 01 Jan 2025 00:00:00 GMT http://hdl.handle.net/123456789/21646 2025-01-01T00:00:00Z http://hdl.handle.net/123456789/21641 Early Retirement: An Open World Game Muhammad Raad Shaukat, 01-136212-031; Muhammad Asif, 01-136212-024 Early Retirement: An Open-World Game with Live NPC Interaction presents an open-world PC game developed in Unity, featuring real-time, voice-activated NPC interactions powered by a fine-tuned compact language model (SmolLM2-360M), which explores themes of trauma and self-discovery, aiming to enhance immersion by replacing static dialogue trees with dynamic, AI-driven conversations; this project specifically investigates whether these real-time, voice-driven AI interactions can significantly enhance narrative immersion and player agency in such open world games. The system utilizes speech processing APIs alongside a custom backend to enable spontaneous, emotionally aware NPC responses to live player speech, a feat achieved through efficient fine-tuning of the language model using a substantial domain-specific dataset of emotional dialogue scenarios, ensuring functionality even under resource constraints. With an average response latency of approximately 3 seconds, the system demonstrates practical viability despite occasional incoherent outputs, and preliminary user testing indicated increased engagement and perceived realism, although further comprehensive evaluation is needed to fully assess long-term usability and sustained coherence. This work therefore contributes a novel framework for integrating lightweight language models into real-time game environments, enabling responsive, emotionally aware NPCs without requiring high-end hardware, thereby showcasing the transformative potential of AI in interactive media and providing valuable insights into the unique challenges and innovative solutions encountered during its development. Supervised by Mr. Qazi Haseeb Yousaf Wed, 01 Jan 2025 00:00:00 GMT http://hdl.handle.net/123456789/21641 2025-01-01T00:00:00Z