Academic papers, new techniques, benchmarks, and theoretical findings in AI/LLM security.
This academic paper analyzes the architecture and security vulnerabilities of Android Automotive OS, the operating system used in connected vehicles. The authors classify and examine potential security threats specific to this automotive platform, which is important because cars increasingly rely on software systems that could be targeted by attackers.
VEXGen is a tool that automates the creation of VEX documents (vulnerability exploitability exchange files, which communicate whether known vulnerabilities can actually be exploited in a specific software project). The tool helps software development teams systematically document which vulnerabilities pose real risks to their projects, rather than treating all reported vulnerabilities as equally dangerous.
This academic paper discusses creating a unified framework for making security decisions, particularly in the context of AI and LLM systems. The work, published in Computers & Security in September 2026, appears to address how organizations can systematically approach security choices rather than making isolated decisions.
This research paper examines how to design effective gamified cyber security training (using game elements like points and levels to teach security skills) by studying loss aversion (people's tendency to fear losing what they have more than gaining something new) and fixed rewards. The study explores which incentive structures motivate people to engage with and learn from security training games most effectively.
This is a research publication analyzing infostealer malware (malicious software that steals personal information like passwords and credentials) using data-driven methods to understand who gets infected and why. The study examines patterns in victim data to learn more about how this type of attack works in the real world.
This academic paper presents Trident, a system for identifying which threat actors (APTs, or advanced persistent threats, which are sophisticated hacking groups) are responsible for cyberattacks by analyzing data from multiple sources using knowledge graphs (databases that store relationships between pieces of information). The research addresses the challenge of attribution when threat intelligence comes from different types of sources and formats.
Researchers have developed tools to reverse-engineer Microsemi flash-based FPGAs (field-programmable gate arrays, which are hardware chips that can be reprogrammed after manufacturing). This work demonstrates that the security protections on these devices can be bypassed, potentially allowing attackers to extract proprietary designs or modify hardware behavior. The research highlights vulnerabilities in how Microsemi protects its FPGA configurations.
N/A -- The provided content is a publication title and metadata only, with no substantive technical details, security issue description, or AI/LLM problem discussed.
This academic paper examines how cybersecurity controls (security measures designed to protect systems from attacks) are actually implemented in contracts used by defense organizations when they buy goods and services. The authors present an empirical study (research based on real-world data collection) and propose a governance integration model (a framework for organizing how security rules are managed and enforced) to improve how these security requirements are handled in procurement.
This research paper proposes a method for creating an accurate software inventory by analyzing what a program actually does in memory at runtime, rather than relying on static files. This approach helps detect supply chain attacks (where malicious code enters software through its dependencies) by catching discrepancies between what the software claims to contain and what it really executes.
This academic article discusses how small and medium-sized enterprises (SMEs, or smaller companies) can manage data privacy using a customized version of the NIST privacy framework (a set of guidelines developed by the U.S. National Institute of Standards and Technology for protecting personal information). The research provides a tailored approach that helps SMEs, which often have fewer resources than large corporations, implement privacy protections that fit their specific needs and capabilities.
Researchers have developed an automated method that uses state machines (mathematical models that track different conditions and transitions between them) to analyze whether software vulnerabilities can actually be exploited in practice. This approach aims to improve how security experts understand and prioritize which bugs pose real threats to systems.
This academic paper introduces MS-Workset, a new dataset designed to identify and track worker misbehavior in mobile crowdsensing (a system where many people use their mobile devices to collect data for research or applications). The dataset helps researchers develop better methods to detect when workers submit low-quality or dishonest data in crowdsourcing projects.
This research paper presents a formal framework for identifying and preventing IMSI catcher attacks, which are devices that impersonate legitimate cell towers to intercept mobile phone communications. The paper, published in January 2027, proposes methods to detect these attacks and protect mobile networks from unauthorized surveillance and data interception.
IDORacle is a research system designed to prevent IDOR (insecure direct object references, where a user can access data belonging to another user by manipulating request parameters) vulnerabilities in database-backed applications by using templates to guide how data access should be controlled at the object level. The system helps ensure that applications properly check whether a user has permission to access a specific piece of data before returning it. This research addresses a common security problem in web applications where developers forget to implement proper authorization checks.
Researchers developed a new method for improving mimic honeypots (fake systems designed to attract and study attackers) by using Minimax-Q game theory (a mathematical approach where two players try to optimize their strategies against each other). The method helps honeypots automatically adapt their defenses to better detect and respond to different attack patterns.
This academic paper introduces the STARC model, a framework for understanding how cybersecurity practices are structured and organized in real-world settings. The paper, published in Computers & Security journal, presents a theoretical approach to studying cybersecurity rather than addressing a specific technical vulnerability or problem.
Researchers developed CROP, a method that uses CPE (Common Platform Enumeration, a standardized way to name software products and versions) and node pruning (removing unnecessary data points) to identify which vulnerabilities are most relevant for attack across different software products. The approach helps security teams prioritize which security risks pose the greatest threat by understanding how vulnerabilities connect across the software ecosystem.
Patronus is a research system that uses statistical decision-making to detect volumetric application-layer DoS attacks (large-scale attempts to overwhelm a web service by flooding it with many requests targeting the application itself rather than the network). The system aims to identify these attacks by analyzing patterns in data rather than relying on traditional rule-based methods.
This research paper proposes a trust model for cloud-edge-end systems (distributed computing environments spanning data centers, edge servers, and end devices) that accounts for uncertainty and time-based context changes. The work addresses how to reliably assess whether different parts of a distributed system can be trusted when information is incomplete or changes over time.