I2S Masters/ Doctoral Theses

All students and faculty are welcome to attend the final defense of I2S graduate students completing their M.S. or Ph.D. degrees. Defense notices for M.S./Ph.D. presentations for this year and several previous years are listed below in reverse chronological order.

Students who are nearing the completion of their M.S./Ph.D. research should schedule their final defenses through the EECS graduate office at least THREE WEEKS PRIOR to their presentation date so that there is time to complete the degree requirements check, and post the presentation announcement online.

Upcoming Defense Notices

Sai Rithvik Gundla

Beyond Regression Accuracy: Evaluating Runtime Prediction for Scheduling Input Sensitive Workloads

When & Where:


Eaton Hall, Room 2001B

Degree Type:

MS Thesis Defense

Committee Members:

Hongyang Sun, Chair
Arvin Agah
David Johnson


Abstract

Runtime estimation plays a structural role in reservation-based scheduling for High Performance Computing (HPC) systems, where predicted walltimes directly influence reservation timing, backfilling feasibility, and overall queue dynamics. This raises a fundamental question of whether improved runtime prediction accuracy necessarily translates into improved scheduling performance. In this work, we conduct an empirical study of runtime estimation under EASY Backfilling using an application-driven workload consisting of MRI-based brain segmentation jobs. Despite identical configurations and uniform metadata, runtimes exhibit substantial variability driven by intrinsic input structure. To capture this variability, we develop a feature-driven machine learning (ML) framework that extracts region-wise features from MRI volumes to predict job runtimes without relying on historical execution traces or scheduling metadata. We integrate these ML-derived predictions into an EASY Backfilling scheduler implemented in the Batsim simulation framework. Our results show that regression accuracy alone does not determine scheduling performance. Instead, scheduling performance depends strongly on estimation bias and its effect on reservation timing and runtime exceedances. In particular, mild multiplicative calibration of ML-based runtime estimates stabilizes scheduler behavior and yields consistently competitive performance across workload and system configurations. Comparable performance can also be observed with certain levels of uniform overestimation; however, calibrated ML predictions provide a systematic mechanism to control estimation bias without relying on arbitrary static inflation. In contrast, underestimation consistently leads to severe performance degradation and cascading job terminations. These findings highlight runtime estimation as a structural control input in backfilling-based HPC scheduling and demonstrate the importance of evaluating prediction models jointly with scheduling dynamics rather than through regression metrics alone.

Devin Setiawan

Concept-Driven Interpretability in Graph Neural Networks: Applications in Neuroscientific Connectomics and Clinical Motor Analysis

When & Where:


Eaton Hall, Room 2001B

Degree Type:

MS Thesis Defense

Committee Members:

Sumaiya Shomaji, Chair
Sankha Guria
Han Wang


Abstract

Graph Neural Networks (GNNs) achieve state-of-the-art performance in modeling complex biological and behavioral systems, yet their "black-box" nature limits their utility for scientific discovery and clinical translation. Standard post-hoc explainability methods typically attribute importance to low-level features, such as individual nodes or edges, which often fail to map onto the high-level, domain-specific concepts utilized by experts. To address this gap, this thesis explores diverse methodological strategies for achieving Concept-Level Interpretability in GNNs, demonstrating how deep learning models can be structurally and analytically aligned with expert domain knowledge. This theme is explored through two distinct methodological paradigms applied to critical challenges in neuroscience and clinical psychology. First, we introduce an interpretable-by-design approach for modeling brain structure-function coupling. By employing an ensemble of GNNs conceptually biased via input graph filtering, the model enforces verifiably disentangled node embeddings. This allows for the quantitative testing of specific structural hypotheses, revealing that a minority of strong anatomical connections disproportionately drives functional connectivity predictions. Second, we present a post-hoc conceptual alignment paradigm for quantifying atypical motor signatures in Autism Spectrum Disorder (ASD). Utilizing a Spatio-Temporal Graph Autoencoder (STGCN-AE) trained on normative skeletal data, we establish an unsupervised anomaly detection system. To provide clinical interpretability, the model's reconstruction error is systematically aligned with a library of human-interpretable kinematic features, such as postural sway and limb jerk. Explanatory meta-modeling via XGBoost and SHAP analysis further translates this abstract loss into a multidimensional clinical signature. Together, these applications demonstrate that integrating concept-level interpretability through either architectural design or systematic post-hoc alignment enables GNNs to serve as robust tools for hypothesis testing and clinical assessment.

Moh Absar Rahman

Permissions vs Promises: Assessing Over-privileged Android Apps via Local LLM-based Description Validation

When & Where:


Eaton Hall, Room 2001B

Degree Type:

MS Thesis Defense

Committee Members:

Drew Davidson, Chair
Sankha Guria
David Johnson


Abstract

Android is the most widely adopted mobile operating system, supporting billions of devices and driven by a robust app ecosystem.  Its permission-based security model aims to enforce the Principle of Least Privilege (PoLP), restricting apps to only the permissions it needs.  However, many apps still request excessive permissions, increasing the risk of data leakage and malicious exploitation. Previous research on overprivileged permission has become ineffective due to outdated methods and increasing technical complexity.  The introduction of runtime permissions and scoped storage has made some of the traditional analysis techniques obsolete.  Additionally, developers often are not transparent in explaining the usage of app permissions on the Play Store, misleading users unknowingly and unwillingly granting unnecessary permissions. This combination of overprivilege and poor transparency poses significant security threats to Android users.  Recently, the rise of local large language models (LLMs) has shown promise in various security fields. The main focus of this study is to analyze whether an app is overpriviledged based on app description provided on the Play Store using Local LLM. Finally, we conduct a manual evaluation to validate the LLM’s findings, comparing its results against human-verified response.

Brinley Hull

An Interactive Virtual Pet for Autism Spectrum Disorder Stress Onset Detection & Mitigation

When & Where:


Nichols Hall, Room 317 (Richard K. Moore Conference Room)

Degree Type:

MS Thesis Defense

Committee Members:

Arvin Agah, Chair
Perry Alexander
David Johnson
Sumaiya Shomaji

Abstract

Individuals with Autism Spectrum Disorder (ASD) frequently experience elevated stress and are at higher risk for mood disorders such as anxiety and depression. Sensory over-responsivity, social challenges, and difficulties with emotional recognition and regulation contribute to such heightened stress. This study presents a proof-of-concept system that detects and mitigates stress through interactions with a virtual pet. Designed for young adults with high-functioning autism, and potentially useful for people beyond that group, the system monitors simulated heart rate, skin resistance, body temperature, and environmental sound and light levels. Upon detection of stress or potential triggers, the system alerts the user and offers stress-reduction activities via a virtual pet, including guided deep-breathing exercises and interactive engagement with the virtual companion. Through combining real-time stress detection with interactive interventions on a single platform, the system aims to help autistic individuals recognize and manage stress more effectively.

Harun Khan

Identifying Weight Surgery Attacks in Siamese Networks

When & Where:


Nichols Hall, Room 246 (Executive Conference Room)

Degree Type:

MS Thesis Defense

Committee Members:

Prasad Kulkami, Chair
Alex Bardas
Bo Luo


Abstract

Facial recognition systems increasingly rely on machine learning services, yet they remain vulnerable to cyber-attacks. While traditional adversarial attacks target input data, an underexplored threat comes from weight manipulation attacks, which directly modify model parameters and can compromise deployed systems in cyber-physical settings. This paper investigates defenses against Weight Surgery, a weight manipulation attack that modifies the final linear layer of neural networks to merge or shatter classes without requiring access to training data. We propose a computationally lightweight defense capable of detecting sample pairs affected by Weight Surgery at low false-positive rates. The defense is designed to operate in realistic deployment scenarios, selecting its sensitivity parameter 𝛾 using only benign samples to meet a target false-positive rate. Evaluation on 1000 independently attacked models demonstrates that our method achieves over 95% recall at a target false-positive rate of 0.001. Performance remains strong even under stricter conditions: at FPR = 0.0001, recall is 92.5%, and at 𝛾=0.98, FPR drops to 0.00001 while maintaining 88.9% recall. These results highlight the robustness and practicality of the defense, offering an effective safeguard for neural networks against model-targeted attacks.

Tanvir Hossain

Security Solutions for Zero-Trust Microelectronics Supply Chains

When & Where:


Nichols Hall, Room 246 (Executive Conference Room)

Degree Type:

PhD Comprehensive Defense

Committee Members:

Tamzidul Hoque, Chair
Drew Davidson
Prasad Kulkarni
Heechul Yun
Huijeong Kim

Abstract

Microelectronics supply chains increasingly rely on globally distributed design, fabrication, integration, and deployment processes, making traditional assumptions of trusted hardware inadequate. Security in this setting can be understood through a zero-trust microelectronics supply-chain model, in which neither manufacturing partners nor procured hardware platforms are assumed trustworthy by default. Two complementary threat scenarios are considered in the proposed research. In the first scenario, custom Integrated Circuits (ICs) fabricated through potentially untrusted foundries are examined, where design-for-security protections intended to prevent piracy, overproduction, and intellectual-property theft can themselves become vulnerable to attacks. In this scenario, hardware Trojan-assisted meta-attacks are used to show that such protections can be systematically identified and subverted by fabrication-stage adversaries. In the second scenario, commercial off-the-shelf ICs are considered from the perspective of end users and procurers, where internal design visibility is unavailable and hardware trustworthiness cannot be directly verified. For this setting, runtime-oriented protection mechanisms are developed to safeguard sensitive computation against malicious hardware behavior and side-channel leakage. Building on these two scenarios, a future research direction is outlined for side-channel-driven vulnerability discovery in off-the-shelf devices, motivated by the need to evaluate and test such platforms prior to deployment when no design information is available. The proposed direction explores gray-box security evaluation using power and electromagnetic side-channel analysis to identify anomalous behaviors and potential vulnerabilities in opaque hardware platforms. Together, these directions establish a foundation for analyzing and mitigating security risks across zero-trust microelectronics supply chains.

Krishna Chaitanya Reddy Chitta

A Dynamic Resource Management Framework and Reconfiguration Strategies for Cloud-native Bulk Synchronous Parallel Applications

When & Where:


Eaton Hall, Room 2001B

Degree Type:

MS Thesis Defense

Committee Members:

Hongyang Sun, Chair
David Johnson
Sumaiya Shomaji


Abstract

Many High Performance Computing (HPC) applications following the Bulk Synchronous Parallel (BSP) model are increasingly deployed in cloud-native, multi-tenant container environments such as Kubernetes. Unlike dedicated HPC clusters, these shared platforms introduce resource virtualization and variability, making BSP applications more susceptible to performance fluctuations.

Workload imbalance across supersteps can trigger the straggler effect, where faster tasks wait at synchronization barriers for slower ones, increasing overall execution time. Existing BSP resource management approaches typically assume static workloads and reuse a single configuration throughout execution. However, real-world workloads vary due to dynamic data and system conditions, making static configurations suboptimal. This limitation underscores the need for adaptive resource management strategies that respond to workload changes while considering reconfiguration costs.

To address these limitations, we evaluate a dynamic, data-driven resource management framework tailored for cloud-native BSP applications. This approach integrates workload profiling, time-series forecasting, and predictive performance modeling to estimate task execution behavior under varying workload and resource conditions. The framework explicitly models the trade-off between performance gains achieved through reconfiguration and the associated checkpointing and migration costs incurred during container reallocation. Multiple reconfiguration strategies are evaluated, spanning simple window-based heuristics, dynamic programming methods, and reinforcement learning approaches. Through extensive experimental evaluation, this framework demonstrates up to 24.5% improvement in total execution time compared to a baseline static configuration. Furthermore, we systematically analyze the performance of each strategy under varying workload characteristics, simulation lengths, and checkpoint penalties, and provide guidance on selecting the most appropriate strategy for a given workload environment.

Smriti Pranjal

NoBIAS: Non-coding RNA Base Interaction Annotation using Visual Snapshot

When & Where:


Slawson Hall, Rm 198

Degree Type:

PhD Comprehensive Defense

Committee Members:

Cuncong Zhong, Chair
Sumaiya Shomaji
Hongyang Sun
Zijun Yao
Xiaoqing Wu

Abstract

Non-coding RNAs fold into complex 3D structures that govern their biological functions, with RNA structural motifs (RSMs) serving as conserved building blocks of this architecture.
These motifs are defined by characteristic base-interaction patterns, making accurate identification and classification of RNA interactions essential for understanding RNA structure and function.

Despite their biological importance, accurately identifying and classifying these interactions remains challenging because the available data are highly variable in quality and scarce in quantity. This compromises annotation reliability, hinders the construction of trustworthy ground truth for systematic assessment, and restricts the supply of reliable training examples needed for supervised learning.

To address this, we introduce NoBIAS, the first resolution-aware, integrated machine learning-based suite for annotating base interactions from 3D RNA structures, inspired by human pattern recognition, augmented with structure prediction for data enrichment, and evaluated on a carefully curated, stratified benchmark.

NoBIAS is a hierarchical framework for RNA base-interaction annotation that integrates interaction-specific inductive biases with multimodal representation learning. By combining a convolution-augmented, rule-guided module for stacking interactions with complementary graph and image encoders for pairing interactions, NoBIAS captures both structural priors and local visual cues of RNA base doublets. A performance-calibrated logit fusion scheme then adaptively integrates modality-specific predictions based on local-structural resolution, enabling robust inference across heterogeneous 3D RNA structures.

Evaluation across multiple benchmark tiers: spanning consensus, homolog-supported, and manually verified cases, shows that NoBIAS consistently outperforms existing methods under increasingly challenging conditions. Together, the NoBIAS design and its evaluation framework provide a systematic foundation for robust RNA base-interaction annotation, enabling more reliable analysis of RNA structure under realistic uncertainty.

Past Defense Notices

Dates

Sarah Johnson

Ordering Attestation Protocols

When & Where:


Nichols Hall, Room 246 (Executive Conference Room)

Degree Type:

PhD Comprehensive Defense

Committee Members:

Perry Alexander, Chair
Michael Branicky
Sankha Guria
Emily Witt
Eileen Nutting

Abstract

Remote attestation is a process of obtaining verifiable evidence from a remote party to establish trust. A relying party makes a request of a remote target that responds by executing an attestation protocol producing evidence reflecting the target's system state and meta-evidence reflecting the evidence’s integrity and provenance. This process occurs in the presence of adversaries intent on misleading the relying party to trust a target they should not. This research introduces a robust approach for evaluating and comparing attestation protocols based on their relative resilience against such adversaries. I develop a Rocq-based, formally-verified mathematical model aimed at describing the difficulty for an active adversary to successfully compromise the attestation. The model supports systematically ranking attestation protocols by the level of adversary effort required to produce evidence that does not accurately reflect the target’s state. My work aims to facilitate the selection of a protocol resilient to adversarial attack.

Lohithya Ghanta

Used Car Analytics

When & Where:


Eaton Hall, Room 2001B

Degree Type:

MS Project Defense

Committee Members:

David Johnson, Chair
Morteza Hashemi
Prasad Kulkarni


Abstract

The used car market is characterized by significant pricing variability, making it challenging for buyers and sellers to determine fair vehicle values. To address this, the project applies a machine learning–driven approach to predict used car prices based on real market data extracted from Cars.com. Following extensive data cleaning, feature engineering, and exploratory analysis, several predictive models were developed and evaluated. Among these, the Stacking Regressor demonstrated superior performance, effectively capturing non-linear pricing patterns and achieving the highest accuracy with the lowest prediction error. Key insights indicate that vehicle age and mileage are the primary drivers of price depreciation, while brand and vehicle category exert notable secondary influence. The resulting pricing model provides a data-backed, transparent framework that supports more informed decision-making and promotes fairness and consistency within the used car marketplace.

Rajmal Shaik

A Human-Guided Approach to Context-Aware SQL Generation in Multi-Agent Frameworks

When & Where:


Eaton Hall, Room 2001B

Degree Type:

MS Project Defense

Committee Members:

Dongjie Wang, Chair
Rachel Jarvis
David Johnson


Abstract

Querying information from relational databases often requires proficiency in SQL, creating a steep learning curve for users who lack programming or database management experience. Text-to-SQL systems aim to bridge this gap by automatically converting natural language questions into executable SQL statements. In recent years, multi-agent frameworks have gained traction for this task, as they enable complex query generation to be decomposed into specialized subtasks such as schema selection based on user intent, SQL synthesis, and refinement of SQL queries through execution-based error correction. This work explores the integration of a human feedback component within a multi-agent Text-to-SQL framework. Human input is introduced after the selector agent identifies relevant schemas and tables, offering targeted guidance before SQL generation. The objective is to examine how such feedback can improve the system’s accuracy and contextual understanding of queries. The implementation leverages OpenAI’s GPT-4.1 mini and GPT-4.1 nano models as the underlying language components. The evaluation is carried out using a standard Text-to-SQL benchmark dataset, focusing on key performance metrics such as execution accuracy and validity efficiency scores.

Ashish Adhikari

Towards assessing the security of program binaries

When & Where:


Eaton Hall, Room 2001B

Degree Type:

PhD Comprehensive Defense

Committee Members:

Prasad Kulkarni, Chair
Alex Bardas
Fengjun Li
Bo Luo

Abstract

Software vulnerabilities are widespread, often resulting from coding weaknesses and poor development practices. These vulnerabilities can be exploited by attackers, posing risks to confidentiality, integrity, and availability. To protect themselves, end-users of software may have an interest in knowing whether the software they purchase, and use is secure from potential attacks. Our work is motivated by this need to automatically assess and rate the security properties of binary software.

While many researchers focus on developing techniques and tools to detect and mitigate vulnerabilities in binaries, our approach is different. We aim to determine whether the software has been developed with proper care. Our hypothesis is that software created with meticulous attention to security is less likely to contain exploitable vulnerabilities. As a first step, we examined the current landscape of binary-level vulnerability detection. We categorized critical coding weaknesses in compiled programming languages and conducted a detailed survey comparing static analysis techniques and tools designed to detect these weaknesses. Additionally, we evaluated the effectiveness of open-source CWE detection tools and analyzed their challenges. To further understand their efficacy, we conducted independent assessments using standard benchmarks.

To determine whether software is carefully and securely developed, we propose several techniques. So far, we have used machine learning and deep learning methods to identify the programming language of a binary at the functional level, enabling us to handle complex cases like mixed-language binaries and we assess whether vulnerable regions in the binary are protected with appropriate security mechanisms. Additionally, we explored the feasibility of detecting secure coding practices by examining adherence to SonarQube’s security-related coding conventions.

Next, we investigate whether compiler warnings generated during binary creation are properly addressed. Furthermore, we also aim to optimize the array bounds detection in the program binary. This enhanced array bounds detection will also increase the effectiveness of detecting secure coding conventions that are related to memory safety and buffer overflow vulnerabilities.

Our ultimate goal is to combine these techniques to rate the overall security quality of a given binary software.

Bayn Schrader

Implementation and Analysis of an Efficient Dual-Beam Radar-Communications Technique

When & Where:


Nichols Hall, Room 246 (Executive Conference Room)

Degree Type:

MS Thesis Defense

Committee Members:

Patrick McCormick, Chair
Shannon Blunt
Jonathan Owen


Abstract

Fully digital arrays enable realization of dual-function radar-communications systems which generate multiple simultaneous transmit beams with different modulation structures in different spatial directions. These spatially diverse transmissions are produced by designing the individual wave forms transmitted at each antenna element that combine in the far-field to synthesize the desired modulations at the specified directions. This thesis derives a look-up table (LUT) implementation of the existing Far-Field Radiated Emissions Design (FFRED) optimization framework. This LUT implementation requires a single optimization routine for a set of desired signals, rather than the previous implementation which required pulse-to-pulse optimization, making the LUT approach more efficient. The LUT is generated by representing the waveforms transmitted by each element in the array as a sequence of beamformers, where the LUT contains beamformers based on the phase difference between the desired signal modulations. The globally optimal beamformers, in terms of power efficiency, can be realized via the Lagrange dual problem for most beam locations and powers. The Phase-Attached Radar-Communications (PARC) waveform is selected for the communications waveform alongside a Linear Frequency Modulated (LFM) waveform for the radar signal. A set of FFRED LUTs are then used to simulate a radar transmission to verify the utility of the radar system. The same LUTs are then used to estimate the communications performance of a system with varying levels of the array knowledge uncertainty.

Will Thomas

Static Analysis and Synthesis of Layered Attestation Protocols

When & Where:


Eaton Hall, Room 2001B

Degree Type:

PhD Comprehensive Defense

Committee Members:

Perry Alexander, Chair
Alex Bardas
Drew Davidson
Sankha Guria
Eileen Nutting

Abstract

Trust is a fundamental issue in computer security. Frequently, systems implicitly trust in other
systems, especially if configured by the same administrator. This fallacious reasoning stems from the belief
that systems starting from a known, presumably good, state can be trusted. However, this statement only
holds for boot-time behavior; most non-trivial systems change state over time, and thus runtime behavior is
an important, oft-overlooked aspect of implicit trust in system security.

To address this, attestation was developed, allowing a system to provide evidence of its runtime behavior to a
verifier. This evidence allows a verifier to make an explicit informed decision about the system’s trustworthiness.
As systems grow more complex, scalable attestation mechanisms become increasingly important. To apply
attestation to non-trivial systems, layered attestation was introduced, allowing attestation of individual
components or layers, combined into a unified report about overall system behavior. This approach enables
more granular trust assessments and facilitates attestation in complex, multi-layered architectures. With the
complexity of layered attestation, discerning whether a given protocol is sufficiently measuring a system, is
executable, or if all measurements are properly reported, becomes increasingly challenging.

In this work, we will develop a framework for the static analysis and synthesis of layered attestation protocols,
enabling more robust and adaptable attestation mechanisms for dynamic systems. A key focus will be the
static verification of protocol correctness, ensuring the protocol behaves as intended and provides reliable
evidence of the underlying system state. A type system will be added to the Copland layered attestation
protocol description language to allow basic static checks, and extended static analysis techniques will be
developed to verify more complex properties of protocols for a specific target system. Further, protocol
synthesis will be explored, enabling the automatic generation of correct-by-construction protocols tailored to
system requirements.

David Felton

Optimization and Evaluation of Physical Complementary Radar Waveforms

When & Where:


Nichols Hall, Room 246 (Executive Conference Room)

Degree Type:

PhD Comprehensive Defense

Committee Members:

Shannon Blunt, Chair
Rachel Jarvis
Patrick McCormick
James Stiles
Zsolt Talata

Abstract

In high dynamic-range environments, matched-filter radar performance is often sidelobe-limited with correlation error being fundamentally constrained by the TB of the collective emission. To contend with the regulatory necessity of spectral containment, the gradient-based complementary-FM framework was developed to produce complementary sidelobe cancellation (CSC) after coherently combining responses from distinct pulses from within a pulse-agile emission. In contrast to most complementary subsets, which were discovered via brute force under the notion of phase-coding, these comp-FM waveform subsets achieve CSC while preserving hardware compatibility since they are FM. Although comp-FM addressed a primary limitation of complementary signals (i.e., hardware distortion), CSC hinges on the exact reconstruction of autocorrelation terms to suppress sidelobes, from which optimality is broken for Doppler shifted signals. This work introduces a Doppler-generalized comp-FM (DG-comp-FM) framework that extends the cancellation condition to account for the anticipated unambiguous Doppler span after post-summing. While this framework is developed for use within a combine-before-Doppler processing manner, it can likewise be employed to design an entire coherent processing interval (CPI) to minimize range-sidelobe modulation (RSM) within the radar point-spread-function (PSF), thereby introducing the potential for cognitive operation if sufficient scattering knowledge is available a-priori. 

Some radar systems operate with multiple emitters, as in the case of Multiple-input-multiple-output (MIMO) radar. Whereas a single emitter must contend with the self-inflicted autocorrelation sidelobes, MIMO systems must likewise contend with the cross-correlation with coincident (in time and spectrum) emissions from other emitters. As such, the determination of "orthogonal waveforms" comprises a large portion of research within the MIMO space, with a small majority now recognizing that true orthogonality is not possible for band-limited signals (albeit, with the exclusion of TDMA). The notion of complementary-FM is proposed for exploration within a MIMO context, whereby coherently combining responses can achieve CSC as well as cross-correlation cancellation for a wide Doppler space. By effectively minimizing cross-correlation terms, this enables improved channel separation on receive as well as improved estimation capability due to reduced correlation error. Proposal items include further exploration/characterization of the space, incorporating an explicit spectral.

Jigyas Sharma

SEDPD: Sampling-Enhanced Differentially Private Defense against Backdoor Poisoning Attacks of Image Classification

When & Where:


Nichols Hall, Room 246 (Executive Conference Room)

Degree Type:

MS Thesis Defense

Committee Members:

Han Wang, Chair
Drew Davidson
Dongjie Wang


Abstract

Recent advancements in explainable artificial intelligence (XAI) have brought significant transparency to machine learning by providing interpretable explanations alongside model predictions. However, this transparency has also introduced vulnerabilities, enhancing adversaries’ ability for the model decision processes through explanation-guided attacks. In this paper, we propose a robust, model-agnostic defense framework to mitigate these vulnerabilities by explanations while preserving the utility of XAI. Our framework employs a multinomial sampling approach that perturbs explanation values generated by techniques such as SHAP and LIME. These perturbations ensure differential privacy (DP) bounds, disrupting adversarial attempts to embed malicious triggers while maintaining explanation quality for legitimate users. To validate our defense, we introduce a threat model tailored to image classification tasks. By applying our defense framework, we train models with pixel-sampling strategies that integrate DP guarantees, enhancing robustness against backdoor poisoning attacks with XAI. Extensive experiments on widely used datasets, such as CIFAR-10, MNIST, CIFAR-100 and Imagenette, and models, including ConvMixer and ResNet-50, show that our approach effectively mitigates explanation-guided attacks without compromising the accuracy of the model. We also test our defense performance against other backdoor attacks, which shows our defense framework can detect other type backdoor triggers very well. This work highlights the potential of DP in securing XAI systems and ensures safer deployment of machine learning models in real-world applications.

Dimple Galla

Intelligent Application for Cold Email Generation: Business Outreach

When & Where:


Eaton Hall, Room 2001B

Degree Type:

MS Project Defense

Committee Members:

David Johnson, Chair
Prasad Kulkarni
Dongjie Wang


Abstract

Cold emailing remains an effective strategy for software service companies to improve organizational reach by acquiring clients. Generic emails often fail to get a response.

This project leverages Generative AI to automate the cold email generation. This project is built with the Llama-3.1 model and a Chroma vector database that supports the semantic search of keywords in the job description that matches the project portfolio links of software service companies. The application automatically extracts the technology related job openings for Fortune 500 companies. Users can either select from these extracted job postings or manually enter URL of a job posting, after which the system generates email and sends email upon approval. Advanced techniques like Chain-of-Thought Prompting and Few-Shot Learning were applied to improve the relevance making the email more responsive. This AI driven approach improves engagement and simplifies the business development process for software service companies.

Shahima Kalluvettu Kuzhikkal

Machine Learning Based Predictive Maintenance for Automotive Systems

When & Where:


Eaton Hall, Room 2001B

Degree Type:

MS Project Defense

Committee Members:

David Johnson, Chair
Rachel Jarvis
Prasad Kulkarni
Hongyang Sun

Abstract

Predictive maintenance plays a central role in reducing vehicle downtime and improving operational efficiency by using data-driven methods to classify the condition of automotive engines. Rather than relying on fixed service schedules or reacting to unexpected breakdowns, this approach leverages machine learning to distinguish between healthy and failed engines based on operational data.

In this project, engine telemetry data capturing key parameters such as engine speed, fuel pressure, and coolant temperature was used to train and evaluate several machine learning models, including logistic regression, random forest, k-nearest neighbors, and a neural network. To further enhance predictive performance, ensemble strategies such as soft voting and stacking were applied. The stacking ensemble, which combines the strengths of multiple classifiers through a meta-learning approach, demonstrated particularly effective results.

This classification-based framework demonstrates how data-driven fault detection can enhance automotive maintenance operations. By identifying engine failures more reliably, machine learning enables safer transportation, reduces maintenance costs, and enhances overall vehicle dependability. Beyond individual vehicles, such approaches have broader applications in fleet management, where proactive decision-making can improve service continuity, reduce operational risks, and increase customer satisfaction.