Research

Text-to-image models (AI systems that generate pictures from text descriptions) can be tricked by attackers who fine-tune them (adjust their parameters on new data) to bypass safety protections and create unsafe images. This paper introduces Patronus, a defensive framework that makes these models more resistant to such attacks by using a specially trained safety decoder (a component that processes the model's internal representations) that produces corrupted outputs for unsafe content while preserving normal image generation for safe requests.

This research presents a new authentication protocol (a method for verifying identity between two parties) designed for Vehicle-to-Grid (V2G) communication, which allows electric vehicles to securely exchange energy with charging stations. The protocol uses lattice-based cryptography (a type of math-based encryption resistant to attacks from quantum computers, which are extremely powerful future computers) and was tested for security using both automated tools and manual analysis to ensure it works safely in real-world situations.

This paper introduces LICA, a new security method that can detect Incorrect-Plaintext Attacks (IPA, where someone encrypts wrong data or falsely claims decryption failed) even against adversaries with quantum computers. LICA uses lattice-based cryptography (mathematical problems believed to be hard even for quantum computers) combined with commitments (binding agreements on encryption parameters) to identify which party is acting maliciously, and presents LICADS, a data sharing protocol that catches quantum IPA by having a trusted arbitrator challenge suspicious parties and assign penalties to the dishonest one.

This research proposes a new method for cloud storage that combines data deduplication (removing duplicate copies of files) with integrity auditing (verifying data hasn't been corrupted or altered) while protecting user privacy. The key innovation is using lightweight algebraic operations (simple mathematical calculations) instead of expensive BLS signatures (complex cryptographic signatures), and applying randomized blinding techniques (adding random noise to hide identifying information) to prevent auditors from discovering who owns which files.

Existing AI systems for finding bugs in smart contracts (programs that run on blockchains) struggle because they only learn from known vulnerabilities and use limited types of information about the code. This paper presents Synesthete, a new detection method that combines multiple types of code features (text, graph structures, and images from different code representations) and uses self-learning to better identify both known and previously unseen vulnerabilities in smart contracts.

DiEL is a method for improving face images while keeping the person's identity recognizable, especially in difficult conditions like extreme angles, blurriness, or poor lighting. The approach uses evolutionary learning (a technique that evolves solutions over time) to separate identity information from pose information (head angle), then reconstructs faces using a pose dictionary (a library of standard face angles learned from many images) to maintain consistency. The method outperforms existing approaches by an average of 4.66% on six benchmark datasets, with particularly strong improvements on challenging cross-pose tests.

This research presents SCVLM (Sparsity-Controllable Vision-Language Model), a system that detects unusual events in videos by learning what normal behavior looks like from unlabeled data, rather than requiring rare examples of anomalies. The system combines vision (image) and language (text) understanding to identify anomalies as deviations from learned normal patterns, while explaining its decisions in a way that matches human reasoning.

Autonomous driving systems (ADS, self-driving car software) often have logical flaws (errors in decision-making rather than traditional bugs) that cause safety accidents and traffic violations, but there are no reliable automated tools to detect them. This paper introduces LawSentry, a framework that uses large language models (LLMs, AI systems trained to understand and generate text) to automatically convert traffic laws into executable test cases called violation oracles that can detect when an ADS behaves illegally, achieving 95.5% accuracy at identifying traffic violations across simulation platforms.

This paper describes EDT-ABC, a new system for threshold attribute-based anonymous credentials (a type of digital credential that lets multiple independent organizations issue permissions without revealing the user's identity). EDT-ABC improves on previous systems by being faster and more flexible, allowing thresholds (minimum numbers of approvals needed) to be adjusted for different services and supporting selective disclosure (where users can prove they have certain permissions without revealing all their information). The researchers demonstrate their system runs significantly faster than existing solutions, reducing credential issuing time by at least 67% and showing time by at least 29%.

Researchers developed a new detection system for AI-generated images (AIGC, or AI-generated content) that combines three techniques to identify fake images created by generative models (AI systems that can create new images from scratch). The system works by examining noise patterns at the sensor level, identifying unusual patterns in the frequency domain (how colors and patterns repeat), and highlighting suspicious regions in images, allowing it to detect synthetic images even when faced with new or unknown generative models.

ESecDT is a framework that allows multiple parties to train decision trees (machine learning models used to make predictions by sorting data into categories) together while keeping their individual data private. It combines two cryptographic techniques called Function Secret Sharing (FSS, a method where a secret is split into parts that only work together) and Replicated Secret Sharing (RSS, another way to distribute secrets across parties) to reduce the amount of data that must be sent between parties during training while maintaining strong privacy protections.

This research addresses a problem in encrypted traffic analysis (ETA, the process of identifying what data is being sent over the internet by examining encrypted network traffic patterns), where existing machine learning methods require lots of manually labeled training data. The authors propose Models-to-Model Learning (M2ML), a new approach that learns from existing ETA models instead of requiring labeled data, using a large language model to align different models' feature spaces (the variables they measure) and resolve disagreements between them based on credibility.

GraphWave is a new AI framework designed to detect malicious network traffic by analyzing multiple types of data together, including the relationships between attackers and targets over time. Unlike older methods that only look at individual data flows, GraphWave combines graph-based analysis (mapping connections between network entities) with deep learning techniques to better distinguish between legitimate and malicious traffic, even when attackers try to hide their patterns through evasion techniques (methods to avoid detection).

MDIGuess is an AI system that improves targeted password guessing attacks by combining multiple types of information about a user, including their past passwords, personal details (PII, or personally identifiable information), and passwords leaked from other services. The system uses an autoregressive neural framework (a type of AI that predicts passwords one piece at a time based on patterns) and achieves a 35% success rate cracking passwords within 1,000 guesses, significantly outperforming existing attack methods.

This research paper presents a new method for radio frequency fingerprinting (RFF, a technique that identifies wireless devices by analyzing unique hardware distortions in their signals) that works better when signals are received by different types of receivers. The proposed approach uses an information bottleneck (IB, a principle that keeps only the most important information while removing unnecessary details) to learn compact representations of device fingerprints without needing labeled data about which receiver collected the signal, outperforming existing methods on real-world datasets.

SymSyz is a hybrid kernel fuzzing framework that combines kernel fuzzing (automated testing that sends random inputs to find bugs) with symbolic execution (a technique that analyzes code paths mathematically to reach difficult-to-access code). The framework addresses three main problems: inconsistent states between the fuzzer and symbolic engine, overhead from tracking too much symbolic information, and slow constraint solving. SymSyz improves testing speed and bug detection by using snapshot-based environment synchronization, limiting symbolic tracking to specific code origins, and prioritizing constraint solving.

Distributed Learning (DL, training AI models across multiple computers) faces problems when some computers are slow (stragglers) or malicious (Byzantine nodes, computers that send incorrect data). FVCC is a new framework that uses coded computing (a technique that adds redundancy so missing data can be recovered) with faster decoding and verification methods to make distributed learning more robust and efficient. The system uses a bidirectional two-dimensional ZigZag Decoding algorithm to recover data quickly and Freivalds' algorithm (a lightweight verification method) to detect dishonest computers.

Federated learning (a way to train AI models where data stays on users' devices instead of being sent to a central server) is vulnerable to backdoor attacks (hidden malicious behaviors inserted into models by attackers), which existing defenses cannot fully stop, especially when attackers poison the model over multiple rounds. Researchers propose Fed-CBE, a defense method that uses three techniques: periodically resetting certain model layers, making the model forget incorrect categories using entropy maximization (spreading predictions evenly across wrong answers), and recovering original task performance through knowledge distillation (copying learned knowledge from older model versions), achieving near-zero attack success rates without hurting normal performance.

This research paper presents PACT, a new method for evaluating decision tree models (algorithms that make predictions by asking yes/no questions in a sequence) while protecting privacy. PACT uses additive homomorphic encryption (a type of math that lets computers do calculations on secret, scrambled data without unscrambling it first) to keep both the tree model and the user's data private, while running faster than previous privacy-preserving approaches.