CyberSec Research

OneAPI is an open standard that supports cross-architecture software development with minimal effort from developers. It brings DPC++ and C++ compilers which need to be thoroughly tested to verify their correctness, reliability, and security. Compilers have numerous code flows and optimization features. This process requires developers with deep understanding of the different compiler flows to craft testcases specific to target paths in the compiler. This testcase creation is a time-consuming and costly process. In this paper, we propose a large-language model (LLM)-based compiler fuzzing tool that integrates the concept of retrieval-augmented generation (RAG). This tool automates the testcase generation task and relieves experienced compiler developers from investing time to craft testcase generation patterns. We test our proposed approach on the Intel DPC++/C++ compiler. This compiler compiles SYCL code and allows developers to offload it to different architectures, e.g. GPUs and CPUs from different vendors. Using this tool, we managed to identify 87 SYCL code test cases that lead to output value mismatch or compiler runtime errors when compiled using Intel DPC++ and clang++ compilers and run on different architectures. The testcases and the identified unexpected behaviors of the compilers under test were obtained within only few hours with no prior background on the compiler passes under tests. This tool facilitates efficient compiler fuzzing with reduced developer time requirements via the dynamic testcase creation capability provided by an LLM with RAG.

In this paper, we ask the question of why the quality of commercial software, in terms of security and safety, does not measure up to that of other (durable) consumer goods we have come to expect. We examine this question through the lens of incentives. We argue that the challenge around better quality software is due in no small part to a sequence of misaligned incentives, the most critical of which being that the harm caused by software problems is by and large shouldered by consumers, not developers. This lack of liability means software vendors have every incentive to rush low-quality software onto the market and no incentive to enhance quality control. Within this context, this paper outlines a holistic technical and policy framework we believe is needed to incentivize better and more secure software development. At the heart of the incentive realignment is the concept of software liability. This framework touches on various components, including legal, technical, and financial, that are needed for software liability to work in practice; some currently exist, some will need to be re-imagined or established. This is primarily a market-driven approach that emphasizes voluntary participation but highlights the role appropriate regulation can play. We connect and contrast this with the EU legal environment and discuss what this framework means for open-source software (OSS) development and emerging AI risks. Moreover, we present a CrowdStrike case study complete with a what-if analysis had our proposed framework been in effect. Our intention is very much to stimulate a robust conversation among both researchers and practitioners.

Large Language Models (LLMs) have recently emerged as powerful tools in cybersecurity, offering advanced capabilities in malware detection, generation, and real-time monitoring. Numerous studies have explored their application in cybersecurity, demonstrating their effectiveness in identifying novel malware variants, analyzing malicious code structures, and enhancing automated threat analysis. Several transformer-based architectures and LLM-driven models have been proposed to improve malware analysis, leveraging semantic and structural insights to recognize malicious intent more accurately. This study presents a comprehensive review of LLM-based approaches in malware code analysis, summarizing recent advancements, trends, and methodologies. We examine notable scholarly works to map the research landscape, identify key challenges, and highlight emerging innovations in LLM-driven cybersecurity. Additionally, we emphasize the role of static analysis in malware detection, introduce notable datasets and specialized LLM models, and discuss essential datasets supporting automated malware research. This study serves as a valuable resource for researchers and cybersecurity professionals, offering insights into LLM-powered malware detection and defence strategies while outlining future directions for strengthening cybersecurity resilience.

As frontier AI advances rapidly, understanding its impact on cybersecurity and inherent risks is essential to ensuring safe AI evolution (e.g., guiding risk mitigation and informing policymakers). While some studies review AI applications in cybersecurity, none of them comprehensively discuss AI's future impacts or provide concrete recommendations for navigating its safe and secure usage. This paper presents an in-depth analysis of frontier AI's impact on cybersecurity and establishes a systematic framework for risk assessment and mitigation. To this end, we first define and categorize the marginal risks of frontier AI in cybersecurity and then systemically analyze the current and future impacts of frontier AI in cybersecurity, qualitatively and quantitatively. We also discuss why frontier AI likely benefits attackers more than defenders in the short term from equivalence classes, asymmetry, and economic impact. Next, we explore frontier AI's impact on future software system development, including enabling complex hybrid systems while introducing new risks. Based on our findings, we provide security recommendations, including constructing fine-grained benchmarks for risk assessment, designing AI agents for defenses, building security mechanisms and provable defenses for hybrid systems, enhancing pre-deployment security testing and transparency, and strengthening defenses for users. Finally, we present long-term research questions essential for understanding AI's future impacts and unleashing its defensive capabilities.

The modern software development landscape heavily relies on transitive dependencies. They enable seamless integration of third-party libraries. However, they also introduce security challenges. Transitive vulnerabilities that arise from indirect dependencies expose projects to risks associated with Common Vulnerabilities and Exposures (CVEs). It happens even when direct dependencies remain secure. This paper examines the lifecycle of transitive vulnerabilities in the Maven ecosystem. We employ survival analysis to measure the time projects remain exposed after a CVE is introduced. Using a large dataset of Maven projects, we identify factors that influence the resolution of these vulnerabilities. Our findings offer practical advice on improving dependency management.

Memory leaks are prevalent in various real-world software projects, thereby leading to serious attacks like denial-of-service. Though prior methods for detecting memory leaks made significant advance, they often suffer from low accuracy and weak scalability for testing large and complex programs. In this paper we present LeakGuard, a memory leak detection tool which provides satisfactory balance of accuracy and scalability. For accuracy, LeakGuard analyzes the behaviors of library and developer-defined memory allocation and deallocation functions in a path-sensitive manner and generates function summaries for them in a bottom-up approach. Additionally, we develop a pointer escape analysis technique to model the transfer of pointer ownership. For scalability, LeakGuard examines each function of interest independently by using its function summary and under-constrained symbolic execution technique, which effectively mitigates path explosion problem. Our extensive evaluation on 18 real-world software projects and standard benchmark datasets demonstrates that LeakGuard achieves significant advancements in multiple aspects: it exhibits superior MAD function identification capability compared to Goshawk, outperforms five state-of-the-art methods in defect detection accuracy, and successfully identifies 129 previously undetected memory leak bugs, all of which have been independently verified and confirmed by the respective development teams.

The Common Vulnerabilities and Exposures (CVEs) system is a reference method for documenting publicly known information security weaknesses and exposures. This paper presents a study of the lifetime of CVEs in software projects and the risk factors affecting their existence. The study uses survival analysis to examine how features of programming languages, projects, and CVEs themselves impact the lifetime of CVEs. We suggest avenues for future research to investigate the effect of various factors on the resolution of vulnerabilities.

Malware detection is increasingly challenged by evolving techniques like obfuscation and polymorphism, limiting the effectiveness of traditional methods. Meanwhile, the widespread adoption of software containers has introduced new security challenges, including the growing threat of malicious software injection, where a container, once compromised, can serve as entry point for further cyberattacks. In this work, we address these security issues by introducing a method to identify compromised containers through machine learning analysis of their file systems. We cast the entire software containers into large RGB images via their tarball representations, and propose to use established Convolutional Neural Network architectures on a streaming, patch-based manner. To support our experiments, we release the COSOCO dataset--the first of its kind--containing 3364 large-scale RGB images of benign and compromised software containers at https://huggingface.co/datasets/k3ylabs/cosoco-image-dataset. Our method detects more malware and achieves higher F1 and Recall scores than all individual and ensembles of VirusTotal engines, demonstrating its effectiveness and setting a new standard for identifying malware-compromised software containers.

System operators responsible for protecting software applications remain hesitant to implement cyber deception technology, including methods that place traps to catch attackers, despite its proven benefits. Overcoming their concerns removes a barrier that currently hinders industry adoption of deception technology. Our work introduces deception policy documents to describe deception technology "as code" and pairs them with Koney, a Kubernetes operator, which facilitates the setup, rotation, monitoring, and removal of traps in Kubernetes. We leverage cloud-native technologies, such as service meshes and eBPF, to automatically add traps to containerized software applications, without having access to the source code. We focus specifically on operational properties, such as maintainability, scalability, and simplicity, which we consider essential to accelerate the adoption of cyber deception technology and to facilitate further research on cyber deception.

Cyber supply chain, encompassing digital asserts, software, hardware, has become an essential component of modern Information and Communications Technology (ICT) provisioning. However, the growing inter-dependencies have introduced numerous attack vectors, making supply chains a prime target for exploitation. In particular, advanced persistent threats (APTs) frequently leverage supply chain vulnerabilities (SCVs) as entry points, benefiting from their inherent stealth. Current defense strategies primarly focus on prevention through blockchain for integrity assurance or detection using plain-text source code analysis in open-source software (OSS). However, these approaches overlook scenarios where source code is unavailable and fail to address detection and defense during runtime. To bridge this gap, we propose a novel approach that integrates multi-source data, constructs a comprehensive dynamic provenance graph, and detects APT behavior in real time using temporal graph learning. Given the lack of tailored datasets in both industry and academia, we also aim to simulate a custom dataset by replaying real-world supply chain exploits with multi-source monitoring.

Current cellular networking remains vulnerable to malicious fake base stations due to the lack of base station authentication mechanism or even a key to enable authentication. We design and build a base station certificate (certifying the base station's public key and location) and a multi-factor authentication (making use of the certificate and the information transmitted in the online radio control communications) to secure the authenticity and message integrity of the base station control communications. We advance beyond the state-of-the-art research by introducing greater authentication factors (and analyzing their individual security properties and benefits), and by using blockchain to deliver the base station digital certificate offline (enabling greater key length or security strength and computational or networking efficiency). We design the certificate construction, delivery, and the multi-factor authentication use on the user equipment. The user verification involves multiple factors verified through the ledger database, the location sensing (GPS in our implementation), and the cryptographic signature verification of the cellular control communication (SIB1 broadcasting). We analyze our scheme's security, performance, and the fit to the existing standardized networking protocols. Our work involves the implementation of building on X.509 certificate (adapted), smart contract-based blockchain, 5G-standardized RRC control communications, and software-defined radios. Our analyses show that our scheme effectively defends against more security threats and can enable stronger security, i.e., ECDSA with greater key lengths. Furthermore, our scheme enables computing and energy to be more than three times efficient than the previous research on the mobile user equipment.

Critical infrastructures integrate a wide range of smart technologies and become highly connected to the cyber world. This is especially true for Cyber-Physical Systems (CPSs), which integrate hardware and software components. Despite the advantages of smart infrastructures, they remain vulnerable to cyberattacks. This work focuses on the cyber resilience of CPSs. We propose a methodology based on knowledge graph modeling and graph analytics to quantify the resilience potential of complex systems by using a multilayered model based on knowledge graphs. Our methodology also allows us to identify critical points. These critical points are components or functions of an architecture that can generate critical failures if attacked. Thus, identifying them can help enhance resilience and avoid cascading effects. We use the SWaT (Secure Water Treatment) testbed as a use case to achieve this objective. This system mimics the actual behavior of a water treatment station in Singapore. We model three resilient designs of SWaT according to our multilayered model. We conduct a resilience assessment based on three relevant metrics used in graph analytics. We compare the results obtained with each metric and discuss their accuracy in identifying critical points. We perform an experimentation analysis based on the knowledge gained by a cyber adversary about the system architecture. We show that the most resilient SWaT design has the necessary potential to bounce back and absorb the attacks. We discuss our results and conclude this work by providing further research axes.

Pairing-based inner product functional encryption provides an efficient theoretical construction for privacy-preserving edge computing secured by widely deployed elliptic curve cryptography. In this work, an efficient software implementation framework for pairing-based function-hiding inner product encryption (FHIPE) is presented using the recently proposed and widely adopted BLS12-381 pairing-friendly elliptic curve. Algorithmic optimizations provide $\approx 2.6 \times$ and $\approx 3.4 \times$ speedup in FHIPE encryption and decryption respectively, and extensive performance analysis is presented using a Raspberry Pi 4B edge device. The proposed optimizations enable this implementation framework to achieve performance and ciphertext size comparable to previous work despite being implemented on an edge device with a slower processor and supporting a curve at much higher security level with a larger prime field. Practical privacy-preserving edge computing applications such as encrypted biomedical sensor data classification and secure wireless fingerprint-based indoor localization are also demonstrated using the proposed implementation framework.

Deep neural networks are not resilient to parameter corruptions: even a single-bitwise error in their parameters in memory can cause an accuracy drop of over 10%, and in the worst cases, up to 99%. This susceptibility poses great challenges in deploying models on computing platforms, where adversaries can induce bit-flips through software or bitwise corruptions may occur naturally. Most prior work addresses this issue with hardware or system-level approaches, such as integrating additional hardware components to verify a model's integrity at inference. However, these methods have not been widely deployed as they require infrastructure or platform-wide modifications. In this paper, we propose a new approach to addressing this issue: training models to be more resilient to bitwise corruptions to their parameters. Our approach, Hessian-aware training, promotes models with $flatter$ loss surfaces. We show that, while there have been training methods, designed to improve generalization through Hessian-based approaches, they do not enhance resilience to parameter corruptions. In contrast, models trained with our method demonstrate increased resilience to parameter corruptions, particularly with a 20$-$50% reduction in the number of bits whose individual flipping leads to a 90$-$100% accuracy drop. Moreover, we show the synergy between ours and existing hardware and system-level defenses.

Zero-knowledge (ZK) protocols enable software developers to provide proofs of their programs' correctness to other parties without revealing the programs themselves. Regular expressions are pervasive in real-world software, and zero-knowledge protocols have been developed in the past for the problem of checking whether an individual string appears in the language of a regular expression, but no existing protocol addresses the more complex PSPACE-complete problem of proving that two regular expressions are equivalent. We introduce Crepe, the first ZK protocol for encoding regular expression equivalence proofs and also the first ZK protocol to target a PSPACE-complete problem. Crepe uses a custom calculus of proof rules based on regular expression derivatives and coinduction, and we introduce a sound and complete algorithm for generating proofs in our format. We test Crepe on a suite of hundreds of regular expression equivalence proofs. Crepe can validate large proofs in only a few seconds each.

Current malware (malicious software) analysis tools focus on detection and family classification but fail to provide clear and actionable narrative insights into the malignant activity of the malware. Therefore, there is a need for a tool that translates raw malware data into human-readable descriptions. Developing such a tool accelerates incident response, reduces malware analysts' cognitive load, and enables individuals having limited technical expertise to understand malicious software behaviour. With this objective, we present MaLAware, which automatically summarizes the full spectrum of malicious activity of malware executables. MaLAware processes Cuckoo Sandbox-generated reports using large language models (LLMs) to correlate malignant activities and generate concise summaries explaining malware behaviour. We evaluate the tool's performance on five open-source LLMs. The evaluation uses the human-written malware behaviour description dataset as ground truth. The model's performance is measured using 11 extensive performance metrics, which boosts the confidence of MaLAware's effectiveness. The current version of the tool, i.e., MaLAware, supports Qwen2.5-7B, Llama2-7B, Llama3.1-8B, Mistral-7B, and Falcon-7B, along with the quantization feature for resource-constrained environments. MaLAware lays a foundation for future research in malware behavior explanation, and its extensive evaluation demonstrates LLMs' ability to narrate malware behavior in an actionable and comprehensive manner.

Supply chain security has become a very important vector to consider when defending against adversary attacks. Due to this, more and more developers are keen on improving their supply chains to make them more robust against future threats. On August 29, 2024 researchers from the Secure Software Supply Chain Center (S3C2) gathered 14 practitioners from 10 government agencies to discuss the state of supply chain security. The goal of the summit is to share insights between companies and developers alike to foster new collaborations and ideas moving forward. Through this meeting, participants were questions on best practices and thoughts how to improve things for the future. In this paper we summarize the responses and discussions of the summit.

The Open Radio Access Network (O-RAN) architecture is reshaping telecommunications by promoting openness, flexibility, and intelligent closed-loop optimization. By decoupling hardware and software and enabling multi-vendor deployments, O-RAN reduces costs, enhances performance, and allows rapid adaptation to new technologies. A key innovation is intelligent network slicing, which partitions networks into isolated slices tailored for specific use cases or quality of service requirements. The RAN Intelligent Controller further optimizes resource allocation, ensuring efficient utilization and improved service quality for user equipment (UEs). However, the modular and dynamic nature of O-RAN expands the threat surface, necessitating advanced security measures to maintain network integrity, confidentiality, and availability. Intrusion detection systems have become essential for identifying and mitigating attacks. This research explores using large language models (LLMs) to generate security recommendations based on the temporal traffic patterns of connected UEs. The paper introduces an LLM-driven intrusion detection framework and demonstrates its efficacy through experimental deployments, comparing non fine-tuned and fine-tuned models for task-specific accuracy.