arXiv subjects

Sayed Erfan Arefin

Publications and source records attributed to Sayed Erfan Arefin.

10 recordsLinked to original sources

arXiv2026

Beyond Pass Rate: A Multilingual, Execution-Grounded Evaluation of Open Code LLMs

Code generation models are typically compared using compact execution benchmarks and aggregate pass rates, but such summaries obscure how performance varies across programming languages, problem families, and failure modes. We present a large-scale, execution-grounded evaluation of 9 openly accessible LLMs specialized for coding on 2,707 free LeetCode problems across 12 programming languages. Our corpus contains 325,343 problem-model-language jobs, each linked to prompt metadata, extracted code, LeetCode execution outcomes, and static-analysis signals. The results show that current open models remain far from the human acceptance reference: the best model, Yi-Coder-9B-Chat, reaches 23.64% mean correctness, compared with a 57.2% human acceptance baseline. Rankings are also slice-dependent: Qwen2.5-Coder-14B-Instruct is strongest on hard problems and distinct-problem coverage, while Gemma-2-27B-IT achieves the highest all-language lint pass rate. Failure analysis shows that compile errors account for 63.25% of non-accepted best submissions, indicating that many failures occur before semantic correctness can be tested. Static quality further diverges from functional correctness. Together, these findings show that multilingual, artifact-preserving evaluation reveals tradeoffs hidden by single-language or single-metric leaderboards.

cs.AI↗

arXiv2026

Privacy-Preserving Covert Communication Using Encrypted Wearable Gesture Recognition

Secure communication is essential in covert and safety-critical settings where verbal interactions may expose user intent or operational context. Wearable gesture-based communication enables low-effort, nonverbal interaction, but existing systems leak motion data, intermediate representations, or inference outputs to untrusted infrastructure, enabling intent inference, behavioral biometric leakage, and insider attacks. This work proposes a privacy-preserving gesture-based covert communication system that ensures, no raw sensor signals, learned features, or classification outputs are exposed to any third-party. The system employs a multi-party homomorphic learning pipeline for gesture recognition directly over encrypted motion data, preventing adversaries from inferring gesture semantics, replaying sensor traces, or accessing intermediate representations. To our knowledge, this work is the first to apply encrypted gesture recognition in a wearable-based covert communication setting. We design and evaluate haptic and visual feedback mechanisms for covert signal delivery and evaluate the system using 600 gesture samples from a commodity smartwatch, achieving over 94.44% classification accuracy and demonstrating the feasibility of the proposed system with practical deployability from high-performance systems to resource-constrained edge devices.

cs.CR↗

arXiv2024

Exploiting HDMI and USB Ports for GPU Side-Channel Insights

Modern computers rely on USB and HDMI ports for connecting external peripherals and display devices. Despite their built-in security measures, these ports remain susceptible to passive power-based side-channel attacks. This paper presents a new class of attacks that exploit power consumption patterns at these ports to infer GPU activities. We develop a custom device that plugs into these ports and demonstrate that its high-resolution power measurements can drive successful inferences about GPU processes, such as neural network computations and video rendering. The ubiquitous presence of USB and HDMI ports allows for discreet placement of the device, and its non-interference with data channels ensures that no security alerts are triggered. Our findings underscore the need to reevaluate and strengthen the current generation of HDMI and USB port security defenses.

cs.CR↗

arXiv2024

Unmasking the giant: A comprehensive evaluation of ChatGPT's proficiency in coding algorithms and data structures

The transformative influence of Large Language Models (LLMs) is profoundly reshaping the Artificial Intelligence (AI) technology domain. Notably, ChatGPT distinguishes itself within these models, demonstrating remarkable performance in multi-turn conversations and exhibiting code proficiency across an array of languages. In this paper, we carry out a comprehensive evaluation of ChatGPT's coding capabilities based on what is to date the largest catalog of coding challenges. Our focus is on the python programming language and problems centered on data structures and algorithms, two topics at the very foundations of Computer Science. We evaluate ChatGPT for its ability to generate correct solutions to the problems fed to it, its code quality, and nature of run-time errors thrown by its code. Where ChatGPT code successfully executes, but fails to solve the problem at hand, we look into patterns in the test cases passed in order to gain some insights into how wrong ChatGPT code is in these kinds of situations. To infer whether ChatGPT might have directly memorized some of the data that was used to train it, we methodically design an experiment to investigate this phenomena. Making comparisons with human performance whenever feasible, we investigate all the above questions from the context of both its underlying learning models (GPT-3.5 and GPT-4), on a vast array sub-topics within the main topics, and on problems having varying degrees of difficulty.

cs.SE↗

arXiv2023

Auditing Lustre file system

With the increasing demand for data storage and the exponential growth of data, traditional single-server architectures are no longer sufficient to handle the massive amounts of data storage, transfer, and various file system events. As a result, distributed file systems have become a necessity to address the scalability challenges of file systems. One such popular distributed file system is Lustre, which is extensively used in high-performance computing environments. Lustre offers parallel file access, allowing multiple clients to access and store data simultaneously. However, in order to ensure the security and integrity of data, auditing plays a crucial role. Lustre auditing serves as a proof of security and enables the implementation of robust security features such as authentication with Kerberos, mandatory access control with SELinux, isolation, and more. Auditing helps track and monitor file system activities, providing valuable insights into user actions, system events, and potential security breaches. The objective of this project is to explore Lustre auditing using CentOS, a popular Linux distribution, within a Lustre architecture. By implementing Lustre auditing, we aim to enhance the security and reliability of the file system. Additionally, we plan to develop a graphical interface that presents the auditing features in a user-friendly and visually appealing manner. This interface will provide administrators and users with a convenient way to monitor and analyze auditing logs, view access patterns, detect anomalies, and ensure compliance with security policies. By combining the power of Lustre's parallel file system architecture with comprehensive auditing capabilities and an intuitive graphical interface, we aim to provide a robust and user-friendly solution for managing and securing large-scale data storage and access.

cs.CR↗

arXiv2023

Simple Two-wheel Self-Balancing Robot Implementation

Cyber-physical systems, also known as CPS, is an emerging field of technology that combines the physical and digital worlds by allowing for seamless interaction and communication between the two. One of the key characteristics of a CPS is its ability to take input from its environment and use that information to produce an output through actuators in the physical world. A balancing robot is a prime example of a CPS, as it uses input from its sensors to continually monitor its orientation and take action to prevent falling over by generating thrust through its wheels or manipulating its inertia. In this specific project, a two-wheel self-balancing robot was developed, utilizing the concept of a reverse pendulum. A reverse pendulum by default is inherently unstable and requires an external force to maintain its balance. In this case, the balancing robot produces this external force through the use of wheels and motors. To achieve precise balancing, stepper motors were utilized in the design of the robot. Additionally, the robot has the capability to move in four basic directions and the movement is controlled through an app connected to the robot via Bluetooth. This allows for remote control and monitoring of the robot's movements and actions. Overall, the development of this two-wheel self-balancing robot serves as a demonstration of the potential and capabilities of cyber-physical systems technology.

cs.RO↗

arXiv2023

Real-life Implementation of Internet of Robotic Things Using 5 DoF Heterogeneous Robotic Arm

Establishing a communication bridge by transferring data driven from different embedded sensors via internet or reconcilable network protocols between enormous number of distinctively addressable objects or "things", is known as the Internet of Things (IoT). IoT can be amalgamated with multitudinous objects such as thermostats, cars, lights, refrigerators, and many more appliances which will be able to build a connection via internet. Where objects of our diurnal life can establish a network connection and get smarter with IoT, robotics can be another aspect which will get beneficial to be brought under the concept of IoT and is able to add a new perception in robotics having "Mechanical Smart Intelligence" which is generally called "Internet of Robotic Things" (IoRT). A robotic arm is a part of robotics where it is usually a programmable mechanical arm which has human arm like functionalities. In this paper, IoRT will be represented by a 5 DoF (degree of freedoms) Robotic Arm which will be able to communicate as an IoRT device, controlled with heterogeneous devices using IoT and "Cloud Robotics".

cs.RO↗

arXiv2023

A Survey on Actionable Knowledge

Actionable Knowledge Discovery (AKD) is a crucial aspect of data mining that is gaining popularity and being applied in a wide range of domains. This is because AKD can extract valuable insights and information, also known as knowledge, from large datasets. The goal of this paper is to examine different research studies that focus on various domains and have different objectives. The paper will review and discuss the methods used in these studies in detail. AKD is a process of identifying and extracting actionable insights from data, which can be used to make informed decisions and improve business outcomes. It is a powerful tool for uncovering patterns and trends in data that can be used for various applications such as customer relationship management, marketing, and fraud detection. The research studies reviewed in this paper will explore different techniques and approaches for AKD in different domains, such as healthcare, finance, and telecommunications. The paper will provide a thorough analysis of the current state of AKD in the field and will review the main methods used by various research studies. Additionally, the paper will evaluate the advantages and disadvantages of each method and will discuss any novel or new solutions presented in the field. Overall, this paper aims to provide a comprehensive overview of the methods and techniques used in AKD and the impact they have on different domains.

cs.LG↗

arXiv2021

Predictive Analysis of Chikungunya

Chikungunya is an emerging threat for health security all over the world which is spreading very fast. Researches for proper forecasting of the incidence rate of chikungunya has been going on in many places in which DARPA has done a very extensive summarized result from 2014 to 2017 with the data of suspected cases, confirmed cases, deaths, population and incidence rate in different countries. In this project, we have analysed the dataset from DARPA and extended it to predict the incidence rate using different features of weather like temperature, humidity, dewiness, wind and pressure along with the latitude and longitude of every country. We had to use different APIs to find out these extra features from 2014-2016. After creating a pure dataset, we have used Linear Regression to predict the incidence rate and calculated the accuracy and error rate.

cs.LG↗

arXiv2021

Second Hand Price Prediction for Tesla Vehicles

The Tesla vehicles became very popular in the car industry as it was affordable in the consumer market and it left no carbon footprint. Due to the large decline in the stock prices of Tesla Inc. at the beginning of 2019, Tesla owners started selling their vehicles in the used car market. These used car prices depended on attributes such as the model of the vehicle, year of production, miles driven, and the battery used for the vehicle. Prices were different for a specific vehicle in different months. In this paper, it is discussed how a machine learning technique is being implemented in order to develop a second-hand Teslavehicle price prediction system. To reach this goal, different machine learning techniques such as decision trees, support vector machine (SVM), random forest, and deep learning were investigated and finally was implemented with boosted decision tree regression. I the future, it is intended to use a more sophisticated algorithm for better accuracy.

cs.LG↗