As the space for digital communication expands, there is even greater need for accessibly simple, transparent, and computationally economical encrypting tools to back up the more operationally challenging ones, like AES and RSA. This article outlines the architecture, implementation and systematic security analysis of a python application with graphical user interface (GUI) in realisation of a hybrid cryptographic pipeline using text cryptography and text transport encoding with a hybrid integrity hashing, the Vigenère polyalphabetic cipher, Base64 encoding and hash generation using SHA-256, respectively. The application is made with Python's GUI library (Tkinter), which makes it run without any external libraries across different operating systems. Four experimental dimensions were considered: (i) the benchmark of computational throughput with classical and modern symmetric cryptographic ciphers; (ii) statistical analysis of Shannon entropy for different categories of messages and Index of Coincidence (IC); (iii) modelling of the sensitivity to key size and resistance to brute force; and (iv) comparative analysis of the multi-dimensional security profile using radar chart. Experimental results demonstrate that for the proposed system keys of length 16 or more yield ciphertext entropy between 5.89 bits/byte and 6.11 bits/byte and IC ranging from 0.037 to 0.041, respectively, closely matching the statistical profile of the statistically random text. In a worst model of 10 attacks per second, theoretical “brute force” time to break the system is > 6 x 10²? years at a key length of 32 characters. The latency of encryption for a 10 KB plaintext is 3.7 ms, which is 3.1-fold overhead over AES-128-GCM — acceptable for the application areas considered. The system is deliberately designed for educational use, low sensitivity inter-system messaging and legacy protocol integration in cases where integration with the AES library is impractical. Formal security restrictions, among other things the lack of IND-CPA guarantees, are characterised and possibilities for mitigation are discussed.

Keywords : Vigenère Cipher, Base64 Encoding, Polyalphabetic Substitution Cipher , Brute-Force Resistance.

Authors : V Hema Sree , P Viswanatha Reddy , M Anjan Kumar , Praveen Naik , Murali Mohan Cheepu

Title : Design and Security Evaluation of a Hybrid Text Encryption Framework Using Vigenère Cipher, Base64 Encoding, and SHA-256 Integrity Verification: A Graphical User Interface Implementation

Volume/Issue : 2025;2(3 ( July - September ))

Page No : 29 - 34

F1 race car bodies and aerodynamics are very much at their limit of performance with laterals exceeding 5g and down force reaching 3.5kN and discernable damage to components being caused by the thermal, mechanical and fatigue effects on the race car at each pit-sting. The traditional telemetry systems are still slower in reacting to faults, detecting faults after the structures go beyond safe operating margins. In this paper, a novel approach that integrates Physics-Informed Neural Networks (PINNs) for real-time state estimation of aerodynamic components into a Digital Twin (DT) computational framework for continuous structural health monitoring (SHM) of motorsport vehicles at high speeds is proposed. The PINN framework uses governing aerodynamics partial differential equations (PDEs) as physics constraint losses in addition to measurement data to provide accurate estimation of lift coefficient C_L, drag coefficient C_D, and composite aerodynamic stability score ? using sparse sensor measurements. The Virtual Vehicle is synchronized with the live vehicle through an Extended Kalman Filter (EKF) and the Structural Health Index (SHI) is calculated per component, and monitored in real time within the Digital Twin. The integrated PINN + Digital Twin framework outperforms CNN, LSTM, sensor fusion, and PINN without any sensors across all the above metrics including fault detection accuracy of 98.5% for 47 hours of vehicle telemetry data over Formula-style driving, physics consistency of 97.2%, detection lead time of 210 ms, and false alarm rate of 0.03/FA/1 hr.

Keywords : PINN, Structural Health Monitoring, Ground-Effect Aerodynamics, Sensor Fusion, Extended Kalman Filter.

Authors : José Varela-Aldás

Title : Physics-Informed Neural Networks and Digital Twin Architecture for Real-Time Aerodynamic State Estimation and Structural Health Monitoring in Formula 1 Vehicles

Volume/Issue : 2025;2(3 ( July - September ))

Page No : 21 - 28

The Large Language Models (LLMs) provides excellent natural language understanding, multi-step reasoning and content creation capabilities. For example, for complex decision making problems, conventional single agent systems have limitations that are intrinsic for the system where everything is done within a single structural constraint planning, memory management, execution and verification. This results in hallucination, failure to keep context in memory at a higher time-scale, failure to decompose task and inadequate adaptability in dynamic environment. The idea is introduced in this paper of the Multi-Agent LLM Framework for Autonomous Decision Systems (MALF-ADS), which fragments the cognitive domains into 4 specialized agents (Planner Agent, Memory Agent, Execution Agent, Evaluation Agent). Agents work independently of each other, have a clear assignment and communicate via structured coordination layer. Incoming tasks are broken down into atomic subtasks by the Planner Agent, persistent contextual knowledge stored in a vector database as part of the Memory Agent, data processed by invoking tools and executing them in the Execution Agent, and validated outputs and computed confidence scores in the Evaluation Agent. The communication protocol in the architecture is dynamic and allows agents to communicate by passing messages which are context-aware. MALF-ADS outperforms all other single LLM, RL agent, and memory-augmented baselines on a 50,000-sample multi-domain dataset in terms of decision accuracy (97%), task success rate (95%), and mean response time (1.3?s).

Keywords : LLM, Agentic AI, Decision Intelligence; Distributed AI, Transformer Architectures, Memory-Augmented Networks.

Authors : Krishna Reddy Atchi

Title : Multi-Agent LLM Framework for Autonomous Decision Systems

Volume/Issue : 2025;2(3 ( July - September ))

Page No : 15 - 20

High-speed motorsport vehicles are exposed to terrific and changing structural loading conditions, up to 340 kmph, and in-situ structural health monitor will be required to continuously monitor the structural condition to avoid component failure and driver safety concerns. Traditional telemetry systems compute the individual values of the sensors and signal only when there is some structural degradation. This paper proposes an architecture for Digital Twin (DT) for real-time structural health monitoring in motorsport vehicle with a continuous synchronization between a virtual replica of the vehicle structure and real vehicle telemetry streams, which enables to predict the degradation of the vehicle components and to estimate the components-based Health Indices, in order to produce predictive maintenance alerts with an average prediction lead time of 210 ms. The proposed architecture includes an LSTM degradation prediction network, an Isolation Forest anomaly detector, a new state synchronization module based on the Kalman filter and a new composite Health Index formulation that is calibrated across front wing, rear wing, suspension, floor and brake components. The Digital Twin framework's anomaly prediction accuracy of 98%, prediction lead time of 210ms, and false positive rate of 4% provide state-of-the-art results, when evaluated against traditional telemetry, LSTM monitoring, and sensor fusion baselines on a 56 lap race simulation dataset when all SHM metrics are considered.

Keywords : Structural Health Monitoring, Sensor Fusion, LSTM, Motorsport Analytics, Kalman Filter, Anomaly Detection.

Authors : Tarun Reddy Nukala

Title : Digital Twin Architecture for Real-Time Structural Health Monitoring in High-Speed Motorsport Vehicles

Volume/Issue : 2025;2(3 ( July - September ))

Page No : 8 - 14

The aerodynamics of modern ground-effect vehicles used in competitive vehicle dynamics applications are highly dynamic and nonlinear, requiring fast, sub-second adaptive aerodynamic control responses. The static control system and single-agent reinforcement learning is not enough to coordinate ride height, wing angle and aerodynamic balance simultaneously at high speed in real time for this vehicle over the entire operational speed range (80-340 km/h). This paper introduces various methods for optimizing aero-control for an autonomously navigating UAV in complex environments, such as suppressing instability and achieving maximum speed in the presence of broad range of disturbances, while maintaining required stability. The paper proposes a novel Multi-Agent Reinforcement Learning (MARL) framework in which three agents (actuation of ride height, aero-control optimization, and aerodynamic balance) will cooperate in sharing a reward signal balancing stability, speed, and instability suppression. The proposed architecture demonstrates that its stability accuracy reaches 97% while its control response time is less than 30ms, outperforms static, single-agent and LSTM-based adaptive architectures on all objective metrics and minimizes energy cost among all the architectures compared in the simulations.

Keywords : Multi-Agent Reinforcement Learning; Ground-Effect Vehicles, Actor-Critic; DDPG; Cooperative Reward; Motorsport AI.

Authors : Vijay Kant

Title : Multi-Agent Reinforcement Learning for Adaptive Aero-Control Optimization in Ground-Effect Vehicles

Volume/Issue : 2025;2(3 ( July - September ))

Page No : 1 - 7