Research

OWASP has released a Top 10 list of security risks specifically for agentic AI applications, which are autonomous AI systems that can use tools and take actions on their own. This framework was built from real incidents and industry experience to help organizations secure these advanced AI systems as they become more common.

The OWASP GenAI Security Project (an open-source community focused on AI safety) has released a list of the top 10 security risks for agentic AI (AI systems that can take actions independently). This guidance was created with input from over 100 industry experts and is meant to help organizations understand and address threats to AI systems.

Deep learning attacks have successfully cracked CAPTCHAs (automated tests that distinguish humans from bots) that use large character sets, especially those with alphabets from languages like Chinese. This paper proposes ACG (Adversarial Large Character Set CAPTCHA Generation), a framework that makes CAPTCHAs harder to attack by adding adversarial perturbations (intentional distortions that confuse AI recognition systems) through two modules: one that prevents character recognition and another that adds global visual noise, reducing attack success rates from 51.52% to 2.56%.

Advanced Persistent Threats (APTs, which are long-term targeted attacks by sophisticated adversaries) are becoming harder to detect early. This paper introduces Warning-Graph, a framework that uses threat intelligence modeling (analyzing data about known attack patterns and infrastructure) to identify ongoing APT attacks by examining IoCs (indicators of compromise, or digital clues that show an attack happened) without needing lots of labeled training data. The framework uses graph-based machine learning techniques to improve detection accuracy by 3-5 percentage points compared to existing methods.

Federated learning (a system where decentralized devices train a shared AI model together while keeping their data local) is vulnerable to poisoning attacks, where malicious participants inject false data to corrupt the final model. This paper proposes AdaptiveShield, a defense system that uses dynamic detection strategies to identify attackers, automatically adjusts its sensitivity thresholds to handle different attack types, reduces damage from missed attackers by adjusting hyperparameters (settings that control how the model learns), and hides user identities through a shuffling mechanism to protect privacy.

This academic paper proposes a new authentication scheme for vehicle-to-grid (V2G) systems, which allow electric vehicles to exchange power with electrical grids. The scheme uses conditional anonymous authentication (a method that hides vehicle identity while allowing identification of bad actors) with a multi-level architecture combining group signatures (cryptographic signatures that hide individual identity within a group) and proxy signatures (where one party can create signatures on behalf of another), making it more efficient than existing approaches.

This paper proposes Verifiable Data Capsule (VDC), a method for secure data sharing in cloud computing where data owners encrypt their data and upload it with access policies to a cloud server, allowing only authorized users to process the data in a TEE (Trusted Execution Environment, a secure zone on a computer where data stays protected). The system addresses a problem with existing approaches: malicious servers could trick users by providing outdated or corrupted data, so the researchers designed a lightweight verification method called Locally Verifiable Chameleon Tag (LVCT) that lets users confirm data hasn't been tampered with or replaced.

One-Time Passwords (OTPs, temporary codes used in two-factor authentication to verify your identity) like HOTP and TOTP have vulnerabilities that let attackers bypass security if they steal the secret key stored on a device or server. This paper proposes HP-OTP, a new OTP scheme that combines your password with the device's unique identifier to make it harder for attackers to forge codes even if they compromise either the device or server.

This research paper describes a method to improve emotion recognition using EEG (electroencephalography, a technology that measures electrical activity in the brain) by generating synthetic EEG data through a diffusion model (a type of AI that creates new data by gradually removing noise from random data). The proposed approach achieved up to 5.6% better accuracy in identifying emotions compared to traditional methods, helping address the problem of not having enough real EEG data for training these systems.

AI systems are valuable for cybersecurity because they can detect patterns and anomalies in large amounts of data, but attackers can exploit these same AI capabilities to launch sophisticated attacks. Adversarial learning (using AI to trick or attack other AI systems) works in two ways: attackers use techniques like data poisoning (corrupting training data) and test time evasion (fooling a trained model with specially crafted inputs) to compromise security systems, while defenders use adversarial training (teaching AI to resist such attacks) to protect against these threats. The source identifies gaps in current research, including a lack of real-world attack data and limited evaluation of AI solutions for network traffic analysis.

This research proposes an AI-based system that uses deep learning and reinforcement learning (RL, a machine learning approach where an AI learns by receiving rewards for good decisions) to detect disease markers in exhaled breath by analyzing volatile organic compounds (VOCs, small carbon-based chemicals produced by the body). The system is designed to work well even with small datasets and aims to improve early disease detection, particularly for chronic kidney disease, through a noninvasive and cost-effective diagnostic method.

Spiking federated learning (FL, a distributed training method where multiple devices collaborate without sharing their private data) typically uses random selection to choose which devices contribute to the final model, but this ignores statistical heterogeneity (differences in data distribution across devices). This paper proposes SFedCA, a new strategy that assigns credits to devices based on their firing intensity (activity level in spiking neural networks, which use brain-inspired neurons that only activate when needed) before and after training, allowing better selection of devices whose data distributions match the overall model needs.

This research paper addresses generalized out-of-distribution detection (OOD detection, where an AI system identifies inputs that are very different from its training data), which is important for AI systems used in safety-critical applications. Rather than focusing on designing better scoring functions, the authors propose a new decision rule called the generalized Benjamini Hochberg procedure that uses hypothesis testing (a statistical method for making decisions about data) to determine whether an input is out-of-distribution, and they prove this method controls false positive rates better than traditional threshold-based approaches.

This paper presents Test-Time Correction (TTC), a system that helps autonomous vehicles fix detection errors while driving, rather than waiting for retraining. TTC uses an Online Adapter module with visual prompts (image-based descriptions of objects derived from feedback like mismatches or user clicks) to continuously correct mistakes in real-time, allowing vehicles to adapt to new situations and improve safety without stopping to retrain the system.

Researchers developed a new method for backdoor attacks (techniques that manipulate AI systems to behave in specific ways when exposed to hidden trigger patterns) that works better in real-world physical scenarios. The method, called VSSC triggers (Visible, Semantic, Sample-specific, and Compatible), uses large language models, generative models, and vision-language models in an automated pipeline to create stealthy triggers that can survive visual distortions and be deployed using real objects, making physical backdoor attacks more practical and systematic than manual methods.

Kubernetes (K8s, a system that manages containerized applications across multiple computers) has a vulnerability in how it handles container image downloads through the CRI-API (the interface between Kubernetes and container runtimes). Because Kubernetes cannot monitor the status of these downloads, attackers can exploit this to launch denial-of-service attacks that consume up to 95% of CPU usage and exhaust network and storage resources on worker nodes indefinitely.

This research paper presents KGEES, a system designed to reduce energy consumption in multi-access edge computing (MEC, a technology that brings servers closer to users for faster processing) while protecting user privacy. The system uses k-anonymity geo-obfuscation (a technique that hides exact user locations by grouping them with others) to keep user locations private, while using a greedy algorithm (an approach that makes quick decisions based on immediate benefits) to decide how to allocate computing resources efficiently.

This research proposes a new framework for side-channel analysis (SCA, a type of attack that exploits physical information like power consumption or timing to break cryptography) by combining multiple different leakage models (ways of measuring how a cryptographic device leaks secrets) using ensemble learning (combining many weaker models into one stronger one). The framework improves how well attackers can recover secret keys by using deep learning with complementary information from different measurement approaches, and the authors prove mathematically that their ensemble model gets closer to the true secret distribution.

Remote attestation (RA, the process of verifying that software running on a trusted computer processor is genuine and hasn't been tampered with) traditionally relies on a single central authority to verify trust, which creates security vulnerabilities. This paper introduces Janus, a new RA system that spreads trust across multiple parties using physical hardware features (PUF, or physically unclonable function, unique identifiers built into computer chips) and smart contracts (automated programs running on blockchain networks) to make the verification process more secure, flexible, and resistant to attacks.