Ashish Kundu

Xin Jin, Charalampos Katsis, Fan Sang et al.

Edge computing is a paradigm that shifts data processing services to the network edge, where data are generated. While such an architecture provides faster processing and response, among other benefits, it also raises critical security issues and challenges that must be addressed. This paper discusses the security threats and vulnerabilities emerging from the edge network architecture spanning from the hardware layer to the system layer. We further discuss privacy and regulatory compliance challenges in such networks. Finally, we argue the need for a holistic approach to analyze edge network security posture, which must consider knowledge from each layer.

Biplav Srivastava, Kausik Lakkaraju, Tarmo Koppel et al.

Chatbots, the common moniker for collaborative assistants, are Artificial Intelligence (AI) software that enables people to naturally interact with them to get tasks done. Although chatbots have been studied since the dawn of AI, they have particularly caught the imagination of the public and businesses since the launch of easy-to-use and general-purpose Large Language Model-based chatbots like ChatGPT. As businesses look towards chatbots as a potential technology to engage users, who may be end customers, suppliers, or even their own employees, proper testing of chatbots is important to address and mitigate issues of trust related to service or product performance, user satisfaction and long-term unintended consequences for society. This paper reviews current practices for chatbot testing, identifies gaps as open problems in pursuit of user trust, and outlines a path forward.

Ganghua Wang, Xun Xian, Jayanth Srinivasa et al.

The growing dependence on machine learning in real-world applications emphasizes the importance of understanding and ensuring its safety. Backdoor attacks pose a significant security risk due to their stealthy nature and potentially serious consequences. Such attacks involve embedding triggers within a learning model with the intention of causing malicious behavior when an active trigger is present while maintaining regular functionality without it. This paper evaluates the effectiveness of any backdoor attack incorporating a constant trigger, by establishing tight lower and upper boundaries for the performance of the compromised model on both clean and backdoor test data. The developed theory answers a series of fundamental but previously underexplored problems, including (1) what are the determining factors for a backdoor attack's success, (2) what is the direction of the most effective backdoor attack, and (3) when will a human-imperceptible trigger succeed. Our derived understanding applies to both discriminative and generative models. We also demonstrate the theory by conducting experiments using benchmark datasets and state-of-the-art backdoor attack scenarios.

Xun Xian, Ganghua Wang, Xuan Bi et al.

Retrieval-Augmented Generation (RAG) has been empirically shown to enhance the performance of large language models (LLMs) in knowledge-intensive domains such as healthcare, finance, and legal contexts. Given a query, RAG retrieves relevant documents from a corpus and integrates them into the LLMs' generation process. In this study, we investigate the adversarial robustness of RAG, focusing specifically on examining the retrieval system. First, across 225 different setup combinations of corpus, retriever, query, and targeted information, we show that retrieval systems are vulnerable to universal poisoning attacks in medical Q\&A. In such attacks, adversaries generate poisoned documents containing a broad spectrum of targeted information, such as personally identifiable information. When these poisoned documents are inserted into a corpus, they can be accurately retrieved by any users, as long as attacker-specified queries are used. To understand this vulnerability, we discovered that the deviation from the query's embedding to that of the poisoned document tends to follow a pattern in which the high similarity between the poisoned document and the query is retained, thereby enabling precise retrieval. Based on these findings, we develop a new detection-based defense to ensure the safe use of RAG. Through extensive experiments spanning various Q\&A domains, we observed that our proposed method consistently achieves excellent detection rates in nearly all cases.

Zhaorui Yang, Haichao Zhu, Qian Zhang et al.

Fault localization identifies program locations responsible for observed failures. Existing techniques rank suspicious code using syntactic spectra--signals derived from execution structure such as statement coverage, control-flow divergence, or dependency reachability. These signals collapse for semantic bugs, where failing and passing executions follow identical code paths and differ only in whether semantic intent is satisfied. Recent LLM-based approaches introduce semantic reasoning but produce stochastic, unverifiable outputs that cannot be systematically cross-referenced across tests or distinguish root causes from cascading effects. We present SemLoc, a fault localization framework based on structured semantic grounding. SemLoc converts free-form LLM reasoning into a closed intermediate representation that binds each inferred property to a typed program anchor, enabling runtime checking and attribution to program structure. It executes instrumented programs to construct a semantic violation spectrum--a constraint-by-test matrix--from which suspiciousness scores are derived analogously to coverage-based methods. A counterfactual verification step further prunes over-approximate constraints and isolates primary causal violations. We evaluate SemLoc on SemFault-250, a corpus of 250 Python programs with single semantic faults. SemLoc outperforms five coverage-, reduction-, and LLM-based baselines, achieving Top-1 accuracy of 42.8% and Top-3 of 68%, while reducing inspection to 7.6% of executable lines. Counterfactual verification provides an additional 12% accuracy gain and identifies primary causal semantic constraints.

An Luo, Jin Du, Fangqiao Tian et al.

Data science plays a critical role in transforming complex data into actionable insights across numerous domains. Recent developments in large language models (LLMs) have significantly automated data science workflows, but a fundamental question persists: Can these agentic AI systems truly match the performance of human data scientists who routinely leverage domain-specific knowledge? We explore this question by designing a prediction task where a crucial latent variable is hidden in relevant image data instead of tabular features. As a result, agentic AI that generates generic codes for modeling tabular data cannot perform well, while human experts could identify the important hidden variable using domain knowledge. We demonstrate this idea with a synthetic dataset for property insurance. Our experiments show that agentic AI that relies on generic analytics workflow falls short of methods that use domain-specific insights. This highlights a key limitation of the current agentic AI for data science and underscores the need for future research to develop agentic AI systems that can better recognize and incorporate domain knowledge.

Mirza Masfiqur Rahman, Ashish Kundu

Generative models such as large language models are extensively used as code copilots and for whole program generation. However, the programs they generate often have questionable correctness, authenticity and reliability in terms of integration as they might not follow the user requirements, provide incorrect and/or nonsensical outputs, or even contain semantic/syntactic errors - overall known as LLM hallucination. In this work, we present several types of code hallucination. We have generated such hallucinated code manually using large language models. We also present a technique - HallTrigger, in order to demonstrate efficient ways of generating arbitrary code hallucination. Our method leverages 3 different dynamic attributes of LLMs to craft prompts that can successfully trigger hallucinations from models without the need to access model architecture or parameters. Results from popular blackbox models suggest that HallTrigger is indeed effective and the pervasive LLM hallucination have sheer impact on software development.

An Luo, Xun Xian, Jin Du et al.

Large language models (LLMs) have advanced the automation of data science workflows. Yet it remains unclear whether they can critically leverage external domain knowledge as human data scientists do in practice. To answer this question, we introduce AssistedDS (Assisted Data Science), a benchmark designed to systematically evaluate how LLMs handle domain knowledge in tabular prediction tasks. AssistedDS features both synthetic datasets with explicitly known generative mechanisms and real-world Kaggle competitions, each accompanied by curated bundles of helpful and adversarial documents. These documents provide domain-specific insights into data cleaning, feature engineering, and model selection. We assess state-of-the-art LLMs on their ability to discern and apply beneficial versus harmful domain knowledge, evaluating submission validity, information recall, and predictive performance. Our results demonstrate three key findings: (1) LLMs frequently exhibit an uncritical adoption of provided information, significantly impairing their predictive performance when adversarial content is introduced, (2) helpful guidance is often insufficient to counteract the negative influence of adversarial information, and (3) in Kaggle datasets, LLMs often make errors in handling time-series data, applying consistent feature engineering across different folds, and interpreting categorical variables correctly. These findings highlight a substantial gap in current models' ability to critically evaluate and leverage expert knowledge, underscoring an essential research direction for developing more robust, knowledge-aware automated data science systems. Our data and code are publicly available here: https://github.com/jeremyxianx/Assisted-DS

Renascence Tarafder Prapty, Ashish Kundu, Arun Iyengar

As the complexity of modern systems increases, so does the importance of assessing their security posture through effective vulnerability management and threat modeling techniques. One powerful tool in the arsenal of cybersecurity professionals is the attack graph, a representation of all potential attack paths within a system that an adversary might exploit to achieve a certain objective. Traditional methods of generating attack graphs involve expert knowledge, manual curation, and computational algorithms that might not cover the entire threat landscape due to the ever-evolving nature of vulnerabilities and exploits. This paper explores the approach of leveraging large language models (LLMs), such as ChatGPT, to automate the generation of attack graphs by intelligently chaining Common Vulnerabilities and Exposures (CVEs) based on their preconditions and effects. It also shows how to utilize LLMs to create attack graphs from threat reports.

Adrian Shuai Li, Arun Iyengar, Ashish Kundu et al.

Many machine learning and data mining algorithms rely on the assumption that the training and testing data share the same feature space and distribution. However, this assumption may not always hold. For instance, there are situations where we need to classify data in one domain, but we only have sufficient training data available from a different domain. The latter data may follow a distinct distribution. In such cases, successfully transferring knowledge across domains can significantly improve learning performance and reduce the need for extensive data labeling efforts. Transfer learning (TL) has thus emerged as a promising framework to tackle this challenge, particularly in security-related tasks. This paper aims to review the current advancements in utilizing TL techniques for security. The paper includes a discussion of the existing research gaps in applying TL in the security domain, as well as exploring potential future research directions and issues that arise in the context of TL-assisted security solutions.

Xun Xian, Ganghua Wang, Xuan Bi et al.

Safeguarding intellectual property and preventing potential misuse of AI-generated images are of paramount importance. This paper introduces a robust and agile plug-and-play watermark detection framework, dubbed as RAW. As a departure from traditional encoder-decoder methods, which incorporate fixed binary codes as watermarks within latent representations, our approach introduces learnable watermarks directly into the original image data. Subsequently, we employ a classifier that is jointly trained with the watermark to detect the presence of the watermark. The proposed framework is compatible with various generative architectures and supports on-the-fly watermark injection after training. By incorporating state-of-the-art smoothing techniques, we show that the framework provides provable guarantees regarding the false positive rate for misclassifying a watermarked image, even in the presence of certain adversarial attacks targeting watermark removal. Experiments on a diverse range of images generated by state-of-the-art diffusion models reveal substantial performance enhancements compared to existing approaches. For instance, our method demonstrates a notable increase in AUROC, from 0.48 to 0.82, when compared to state-of-the-art approaches in detecting watermarked images under adversarial attacks, while maintaining image quality, as indicated by closely aligned FID and CLIP scores.

Aditya Patwardhan, Vivek Vaidya, Ashish Kundu

Generative AI (Gen AI) with large language models (LLMs) are being widely adopted across the industry, academia and government. Cybersecurity is one of the key sectors where LLMs can be and/or are already being used. There are a number of problems that inhibit the adoption of trustworthy Gen AI and LLMs in cybersecurity and such other critical areas. One of the key challenge to the trustworthiness and reliability of LLMs is: how consistent an LLM is in its responses? In this paper, we have analyzed and developed a formal definition of consistency of responses of LLMs. We have formally defined what is consistency of responses and then develop a framework for consistency evaluation. The paper proposes two approaches to validate consistency: self-validation, and validation across multiple LLMs. We have carried out extensive experiments for several LLMs such as GPT4oMini, GPT3.5, Gemini, Cohere, and Llama3, on a security benchmark consisting of several cybersecurity questions: informational and situational. Our experiments corroborate the fact that even though these LLMs are being considered and/or already being used for several cybersecurity tasks today, they are often inconsistent in their responses, and thus are untrustworthy and unreliable for cybersecurity.

Kaiyuan Zhang, Siyuan Cheng, Hanxi Guo et al.

Large language models (LLMs) have achieved remarkable success and are widely adopted for diverse applications. However, fine-tuning these models often involves private or sensitive information, raising critical privacy concerns. In this work, we conduct the first comprehensive study evaluating the vulnerability of fine-tuned LLMs to membership inference attacks (MIAs). Our empirical analysis demonstrates that MIAs exploit the loss reduction during fine-tuning, making them highly effective in revealing membership information. These findings motivate the development of our defense. We propose SOFT (\textbf{S}elective data \textbf{O}bfuscation in LLM \textbf{F}ine-\textbf{T}uning), a novel defense technique that mitigates privacy leakage by leveraging influential data selection with an adjustable parameter to balance utility preservation and privacy protection. Our extensive experiments span six diverse domains and multiple LLM architectures and scales. Results show that SOFT effectively reduces privacy risks while maintaining competitive model performance, offering a practical and scalable solution to safeguard sensitive information in fine-tuned LLMs.

Pratik Sonune, Ritwik Rai, Shamik Sural et al.

Organizations often lay down rules or guidelines called Natural Language Access Control Policies (NLACPs) for specifying who gets access to which information and when. However, these cannot be directly used in a target access control model like Attribute-based Access Control (ABAC). Manually translating the NLACP rules into Machine Enforceable Security Policies (MESPs) is both time consuming and resource intensive, rendering it infeasible especially for large organizations. Automated machine translation workflows, on the other hand, require information security officers to be adept at using such processes. To effectively address this problem, we have developed a free web-based publicly accessible tool called LMN (LLMs for generating MESPs from NLACPs) that takes an NLACP as input and converts it into a corresponding MESP. Internally, LMN uses the GPT 3.5 API calls and an appropriately chosen prompt. Extensive experiments with different prompts and performance metrics firmly establish the usefulness of LMN.

Christian Rondanini, Barbara Carminati, Elena Ferrari et al.

The rapid evolution of malware attacks calls for the development of innovative detection methods, especially in resource-constrained edge computing. Traditional detection techniques struggle to keep up with modern malware's sophistication and adaptability, prompting a shift towards advanced methodologies like those leveraging Large Language Models (LLMs) for enhanced malware detection. However, deploying LLMs for malware detection directly at edge devices raises several challenges, including ensuring accuracy in constrained environments and addressing edge devices' energy and computational limits. To tackle these challenges, this paper proposes an architecture leveraging lightweight LLMs' strengths while addressing limitations like reduced accuracy and insufficient computational power. To evaluate the effectiveness of the proposed lightweight LLM-based approach for edge computing, we perform an extensive experimental evaluation using several state-of-the-art lightweight LLMs. We test them with several publicly available datasets specifically designed for edge and IoT scenarios and different edge nodes with varying computational power and characteristics.

Xin Jin, Charalampos Katsis, Fan Sang et al.

The rampant occurrence of cybersecurity breaches imposes substantial limitations on the progress of network infrastructures, leading to compromised data, financial losses, potential harm to individuals, and disruptions in essential services. The current security landscape demands the urgent development of a holistic security assessment solution that encompasses vulnerability analysis and investigates the potential exploitation of these vulnerabilities as attack paths. In this paper, we propose Graphene, an advanced system designed to provide a detailed analysis of the security posture of computing infrastructures. Using user-provided information, such as device details and software versions, Graphene performs a comprehensive security assessment. This assessment includes identifying associated vulnerabilities and constructing potential attack graphs that adversaries can exploit. Furthermore, Graphene evaluates the exploitability of these attack paths and quantifies the overall security posture through a scoring mechanism. The system takes a holistic approach by analyzing security layers encompassing hardware, system, network, and cryptography. Furthermore, Graphene delves into the interconnections between these layers, exploring how vulnerabilities in one layer can be leveraged to exploit vulnerabilities in others. In this paper, we present the end-to-end pipeline implemented in Graphene, showcasing the systematic approach adopted for conducting this thorough security analysis.

Đorđe Klisura, Joseph Khoury, Ashish Kundu et al.

Access control is a cornerstone of secure computing, yet large language models often blur role boundaries by producing unrestricted responses. We study role-conditioned refusals, focusing on the LLM's ability to adhere to access control policies by answering when authorized and refusing when not. To evaluate this behavior, we created a novel dataset that extends the Spider and BIRD text-to-SQL datasets, both of which have been modified with realistic PostgreSQL role-based policies at the table and column levels. We compare three designs: (i) zero or few-shot prompting, (ii) a two-step generator-verifier pipeline that checks SQL against policy, and (iii) LoRA fine-tuned models that learn permission awareness directly. Across multiple model families, explicit verification (the two-step framework) improves refusal precision and lowers false permits. At the same time, fine-tuning achieves a stronger balance between safety and utility (i.e., when considering execution accuracy). Longer and more complex policies consistently reduce the reliability of all systems. We release RBAC-augmented datasets and code.

Vivek Vaidya, Aditya Patwardhan, Ashish Kundu

Generative AI technologies, particularly Large Language Models (LLMs), are rapidly being adopted across industry, academia, and government sectors, owing to their remarkable capabilities in natural language processing. However, despite their strengths, the inconsistency and unpredictability of LLM outputs present substantial challenges, especially in security-critical domains such as access control. One critical issue that emerges prominently is the consistency of LLM-generated responses, which is paramount for ensuring secure and reliable operations. In this paper, we study the application of LLMs within the context of Cybersecurity Access Control Systems. Specifically, we investigate the consistency and accuracy of LLM-generated password policies, translating natural language prompts into executable pwquality$.$conf configuration files. Our experimental methodology adopts two distinct approaches: firstly, we utilize pre-trained LLMs to generate configuration files purely from natural language prompts without additional guidance. Secondly, we provide these models with official pwquality$.$conf documentation to serve as an informative baseline. We systematically assess the soundness, accuracy, and consistency of these AI-generated configurations. Our findings underscore significant challenges in the current generation of LLMs and contribute valuable insights into refining the deployment of LLMs in Access Control Systems.

Kausik Lakkaraju, Vedant Khandelwal, Biplav Srivastava et al.

Artificial intelligence (AI) has the potential to transform education with its power of uncovering insights from massive data about student learning patterns. However, ethical and trustworthy concerns of AI have been raised but are unsolved. Prominent ethical issues in high school AI education include data privacy, information leakage, abusive language, and fairness. This paper describes technological components that were built to address ethical and trustworthy concerns in a multi-modal collaborative platform (called ALLURE chatbot) for high school students to collaborate with AI to solve the Rubik's cube. In data privacy, we want to ensure that the informed consent of children, parents, and teachers, is at the center of any data that is managed. Since children are involved, language, whether textual, audio, or visual, is acceptable both from users and AI and the system can steer interaction away from dangerous situations. In information management, we also want to ensure that the system, while learning to improve over time, does not leak information about users from one group to another.

Liou Tang, James Joshi, Ashish Kundu

Machine Unlearning (MU) aims to update Machine Learning (ML) models following requests to remove training samples and their influences on a trained model efficiently without retraining the original ML model from scratch. While MU itself has been employed to provide privacy protection and regulatory compliance, it can also increase the attack surface of the model. Existing privacy inference attacks towards MU that aim to infer properties of the unlearned set rely on the weaker threat model that assumes the attacker has access to both the unlearned model and the original model, limiting their feasibility toward real-life scenarios. We propose a novel privacy attack, A Posteriori Label-Only Membership Inference Attack towards MU, Apollo, that infers whether a data sample has been unlearned, following a strict threat model where an adversary has access to the label-output of the unlearned model only. We demonstrate that our proposed attack, while requiring less access to the target model compared to previous attacks, can achieve relatively high precision on the membership status of the unlearned samples.

Kshitij Gulati, Gaurav Verma, Mukesh Mohania et al.

With the global metamorphosis of the beauty industry and the rising demand for beauty products worldwide, the need for an efficacious makeup recommendation system has never been more. Despite the significant advancements made towards personalised makeup recommendation, the current research still falls short of incorporating the context of occasion in makeup recommendation and integrating feedback for users. In this work, we propose BeautifAI, a novel makeup recommendation system, delivering personalised occasion-oriented makeup recommendations to users while providing real-time previews and continuous feedback. The proposed work's novel contributions, including the incorporation of occasion context, region-wise makeup recommendation, real-time makeup previews and continuous makeup feedback, set our system apart from the current work in makeup recommendation. We also demonstrate our proposed system's efficacy in providing personalised makeup recommendation by conducting a user study.

Ashish Kundu, Arun Ayachitula, Nagamani Sistla

In this paper, we have studied how the text of an ancient literature on how their integrity has been preserved for several centuries. Specifically, The Vedas is an ancient literature, which has its text remained preserved without any corruption for thousands of years. As we studied the system that protects the integrity of the text, pronunciation and semantics of the The Vedas, we discovered a number of similarities it has with the current concept of blockchain technology. It is surprising that the notion of de-centralized trust and mathematical encodings have existed since thousands of years in order to protect this work of literature. We have presented our findings and analysis of the similarities. There are also certain technical mechanisms that The Vedic integrity system uses, which can be used to enhance the current digital blockchain platforms in terms of its security and robustness.

Josh Payne, Ashish Kundu

In cloud computing environments with many virtual machines, containers, and other systems, an epidemic of malware can be highly threatening to business processes. In this vision paper, we introduce a hierarchical approach to performing malware detection and analysis using several recent advances in machine learning on graphs, hypergraphs, and natural language. We analyze individual systems and their logs, inspecting and understanding their behavior with attentional sequence models. Given a feature representation of each system's logs using this procedure, we construct an attributed network of the cloud with systems and other components as vertices and propose an analysis of malware with inductive graph and hypergraph learning models. With this foundation, we consider the multicloud case, in which multiple clouds with differing privacy requirements cooperate against the spread of malware, proposing the use of federated learning to perform inference and training while preserving privacy. Finally, we discuss several open problems that remain in defending cloud computing environments against malware related to designing robust ecosystems, identifying cloud-specific optimization problems for response strategy, action spaces for malware containment and eradication, and developing priors and transfer learning tasks for machine learning models in this area.

Qi Zhang, Zehra Sura, Ashish Kundu et al.

Data exfiltration attacks have led to huge data breaches. Recently, the Equifax attack affected 147M users and a third-party library - Apache Struts - was alleged to be responsible for it. These attacks often exploit the fact that sensitive data are stored unencrypted in process memory and can be accessed by any function executing within the same process, including untrusted third-party library functions. This paper presents StackVault, a kernel-based system to prevent sensitive stack-based data from being accessed in an unauthorized manner by intra-process functions. Stack-based data includes data on stack as well as data pointed to by pointer variables on stack. StackVault consists of three components: (1) a set of programming APIs to allow users to specify which data needs to be protected, (2) a kernel module which uses unforgeable function identities to reliably carry out the sensitive data protection, and (3) an LLVM compiler extension that enables transparent placement of stack protection operations. The StackVault system automatically enforces stack protection through spatial and temporal access monitoring and control over both sensitive stack data and untrusted functions. We implemented StackVault and evaluated it using a number of popular real-world applications, including gRPC. The results show that StackVault is effective and efficient, incurring only up to 2.4% runtime overhead.

Ashish Kundu

In this paper, we have introduced the notion of "Principle of Need-to- Act". This principle is essential towards developing secure systems, security solutions and analyzing security of a solution.

Ashish Kundu, Zhiqiang Lin, Joshua Hammond

All mobile devices are energy-constrained. They use batteries that allows using the device for a limited amount of time. In general, energy attacks on mobile devices are denial of service (DoS) type of attacks. While previous studies have analyzed the energy attacks in servers, no existing work has analyzed the energy attacks on mobile devices. As such, in this paper, we present the first systematic study on how to exploit the energy attacks on smartphones. In particular, we explore energy attacks from the following aspect: hardware components, software resources, and network communications through the design and implementation of concrete malicious apps, and malicious web pages. We quantitatively show how quickly we can drain the battery through each individual attack, as well as their combinations. Finally, we believe energy exploit will be a practical attack vector and mobile users should be aware of this type of attacks.

Ashish Kundu, Chinmay Kundu, Karan K. Budhraja

Given that security threats and privacy breaches are com- monplace today, it is an important problem for one to know whether their device(s) are in a "good state of security", or is there a set of high- risk vulnerabilities that need to be addressed. In this paper, we address this simple yet challenging problem. Instead of gaining white-box access to the device, which offers privacy and other system issues, we rely on network logs and events collected offine as well as in realtime. Our approach is to apply analytics and machine learning for network security analysis as well as analysis of the security of the overall device - apps, the OS and the data on the device. We propose techniques based on analytics in order to determine sensitivity of the device, vulnerability rank of apps and of the device, degree of compromise of apps and of the device, as well as how to define the state of security of the device based on these metrics. Such metrics can be used further in machine learning models in order to predict the users of the device of high risk states, and how to avoid such risks.