1393 research outputs found
Sort by
Evaluating Large Language Models for Enhanced Fuzzing: An Analysis Framework for LLM-Driven Seed Generation
No full textFuzzing is a crucial technique for detecting software defects by dynamically generating and testing program inputs. This study introduces a framework designed to assess the application of Large Language Models (LLMs) to automate the generation of effective seed inputs for fuzzing, particularly in the Python programming environment where traditional approaches are less effective. Utilizing the Atheris fuzzing framework, we created over 38,000 seed inputs from LLMs targeted at 50 Python functions from widely-used libraries. Our findings underscore the critical role of LLM selection in seed effectiveness. In certain cases, seeds generated by LLMs rivaled or surpassed traditional fuzzing campaigns, with a corpus of fewer than 100 LLM-generated entries outperforming over 100,000 conventionally produced inputs. These seeds significantly improved code coverage and instruction count during fuzzing sessions, illustrating the efficacy of our framework in facilitating an automated, scalable approach to evaluating LLM effectiveness. The results, validated through linear regression analysis, demonstrate that selecting the appropriate LLM based on its training and capabilities is essential for optimizing fuzzing efficiency and facilitates the testing of future LLM versions
Automatic Heap Primitive Detection for Exploit Development
Get PDFThe problem of fully automating heap exploit generation remains unsolved. An examination of current research in this field reveals that it is both a challenging research topic and one that is currently understudied relative to stack-based automation. This research has the potential to help understand and detect memory-related issues, such as memory leaks, within a program, without needing to spend significant time studying the program or to more quickly triage a given vulnerability to understand its potential impacts and the possibility for exploitation. Even in the most compelling related research, there are significant gaps before it will be possible to fully automate heap-based exploit development in a realistic environment. To achieve the research goal of decreasing the time required to perform heap exploitation by developing a tool to accurately detect all heap primitives within a target program in an automated fashion, an artifact to detect these heap primitives must be created. This artifact will be evaluated based on a set of requirements at the completion of the research. The first of these requirements is that the analysis can be completed within bounded time, avoiding issues such as infinite loops or state explosion, which are common issues with symbolic execution. Secondly, the artifact should perform an exhaustive search for all primitives within the target software, even if this results in false positives. Additionally, the artifact must provide associated details about the primitives that it is able to detect, including the code path for the primitive and other allocations that occur as a result of the code in the primitive. Finally, accuracy of the artifact is of the highest importance, ensuring that correct details about the detected primitives are reported
Toward an Insider Threat Education Platform: A Theoretical Literature Review
Get PDFInsider threats (InTs) within organizations are small in number but have a disproportionate ability to damage systems, information, and infrastructure. Existing InT research studies the problem from psychological, technical, and educational perspectives. Proposed theories include research on psychological indicators, machine learning, user behavioral log analysis, and educational methods to teach employees recognition and mitigation techniques. Because InTs are a human problem, training methods that address InT detection from a behavioral perspective are critical. While numerous technological and psychological theories exist on detection, prevention, and mitigation, few training methods prioritize psychological indicators. This literature review studied peer-reviewed, InT research organized by subtopic and extracted critical theories from psychological, technical, and educational disciplines. In doing so, this is the first study to comprehensively organize research across all three approaches in a manner which properly informs the development of an InT education platform
Collectivism, Individualism, and Cyber Crime: An Evaluation of the United States, the Netherlands, China, and Pakistan
Get PDFScholars have exhaustively examined the larger role social and cultural forces play in determining and shaping criminal behavior. Many of these scholars have examined a dichotomy that sees modern states as falling into either a culture of broad collectivism or one of clear individualism. The relatively recent emergence of cybercrime, however, raises the question of whether this dichotomy affects a propensity for internalized cybercrime. In focusing on four states, China and Pakistan as generally accepted collectivist cultures and the United States and the Netherlands as generally accepted individualist cultures, this research seeks to examine the extent to which, if at all, such values affect rates of internal cybercrime. In determining whether the individualism/collectivism dichotomy applies to frequency of internal cybercrime, this research seeks to determine if existing criminological research applies based on cultural backgrounds or if researchers need to consider cybercrime as entirely distinct, requiring new theories and frameworks in which to understand and combat it
Micromouse - Maze Solving Robot
Get PDFA Micromouse is a path-finding robot designed to traverse and map a maze to determine the shortest path possible. These robots are typically entered in annual Micromouse competitions to compete against other teams for achieving the best possible time in solving the maze in the most efficient manner.https://scholar.dsu.edu/research-symposium/1031/thumbnail.jp
Resonant Landscapes
Get PDFImmersive ambisonics field recordings of 13 South Dakota State Parks. Used for: interactive smartphone-based sound art exhibit and immersive soundscape compositions. Goal: by studying a park\u27s soundscape through immersive recording, we can gain a deeper understanding and appreciation for its unique sonic characteristics.https://scholar.dsu.edu/research-symposium/1037/thumbnail.jp
Towards Robust IoT Security: A Blockchain Design with Attribute-based Encryption
Get PDFInternet of Things (IoT) data and devices face significant security concerns. It is well-known that IoT security standards and frameworks have yet to be established due to variations with IoT developers and manufacturers agreeing on fundamental design methods to address IoT security and its limited computational resources. This research aims to determine how security standards and frameworks can be better implemented to enhance IoT security. Specifically, this research investigates whether implementing an encryption scheme such as attribute-based encryption (ABE) along with blockchain, a distributed ledger technology (DLT), strengthens the security of IoT data and devices. In this context, an encryption scheme is applied at the IoT gateway on sensor data while implementing blockchain for non-repudiation of data as it traverses the network infrastructure. Finally, we present a performance analysis to assess our proposed solution and validate the performance of ciphertext-policy attribute-based encryption (CP-ABE) based on the number of attributes used in the access policy
Implementing a Zero Trust Architecture For ICS/SCADA Systems
Get PDFThe American people depend significantly on Industrial Control Systems (ICS), which help enable critical infrastructures worldwide. These systems can deliver vital infrastructure services to provide essential capabilities worldwide. These Industrial Control Systems also contain many associated vulnerabilities that go unmitigated due to the complex nature of the systems. To strengthen and improve that cyber posture, a more proactive cyber approach utilizing a Zero Trust Architecture was analyzed as a solution to mitigate the risk. The ICS Purdue Model was interpreted as a potential framework for ICS architecture during the analysis. The ICS Purdue Model was then enhanced with Zero Trust controls as a possible cyber solution to mitigate the associated risks. Many of these systems are complex and have to maintain 24/7/365 operations. Because of the unique nature of these systems, they cannot be upgraded, or patched with modern-day cyber controls. This design science research study will investigate an Enhanced ICS Purdue Model with Zero Trust controls as an alternative approach to mitigate ICS cyber risks, rather than traditional cyber defense enterprise IT approaches
Weed Detection using Lightweight DL models with Transfer Learning & Hyperparameter Optimization
No full textWeeds significantly reduce agricultural yields. This study focuses on using artificial intelligence, specifically deep learning, for early-stage weed detection on edge devices, which have limited computational power. We propose a three-part strategy to address this challenge: employing lightweight architectures to reduce the model size and computational demand; using transfer learning to overcome the limitations of small data sets; and applying Bayesian optimization to fine-tune model parameters. Our results show that MobileNetV2 and EfficientNetV2B0 models achieve high accuracy (95.09% and 95.79% respectively) with MobileNetV2 being nearly as accurate but much smaller in size (13.96 MB compared to 23 MB). This demonstrates MobileNetV2\u27s suitability for computationally constrained edge devices
AI Capability and Supply Chain Performance: A Systematic Literature Review
No full textSupply chain management (SCM) faces challenges such as uncertainty, volatility, and disruption. Artificial intelligence (AI) can improve SCM by enabling data-driven decision-making, automation, and optimization. However, the literature on AI in SCM is fragmented and a systematic review is lacking. This study evaluated the effectiveness of AI capabilities on supply chain performance through a systematic literature review (SLR) of empirical studies. SLR follows PRISMA guidelines and answers three research questions: (1) What are the types and amounts of AI capabilities in SCM? (2) How do AI capabilities affect supply chain performance? (3) What factors moderate or mediate the relationship between artificial intelligence and performance? SLR searches large databases for relevant articles and uses a thematic analysis approach to synthesize the data. The study explored the application of artificial intelligence in SCM, identified gaps and limitations in the literature, and provided implications and recommendations for managers