CCS 2025

160 papers
CCS 2025 featured 155 papers with a strong emphasis on the security and privacy of machine learning models. Research themes included adversarial attacks on multi-modal sensor fusion, temporal consistency in video manipulation, and the poisoning of text-to-image models via adversarial mislabeling. Privacy auditing was also a central focus, with studies examining membership inference for foundation models and the concentration of privacy leakage in data tails. Defense mechanisms, such as surgical neuron pruning and weight space analysis for backdoor detection, were frequently discussed. Beyond AI security, the conference addressed system-level vulnerabilities and cryptographic improvements. Investigations covered implicit permissions in Kubernetes, kernel-level isolation on AArch64, and automated vulnerability hunting in Windows ALPC. Other research explored machine unlearning in generative models and modular machine learning for organic chemistry. Prolific contributors to this edition included Shuai Wang, Tianwei Zhang, Daoyuan Wu, Ke Xu, and Qi Li.

Top Research Directions

Research directions most represented by CCS 2025 papers.

Knowledge Poisoning in Retrieval Augmented Generation 3
The Vulnerability of Static Defenses to Adaptive Adversaries 3
Data Efficient Backdoor Purification for Neural Networks 2
Scaling Membership Inference Beyond Shadow Model Ensembles 2
Secure Computation with Hardware Enclaves 2
Information Leakage from Encrypted Traffic Metadata 2
Mitigating the Alignment Tax in Large Language Models 2
The Mechanics of Backdoor Attacks in Neural Networks 2
Distributed Backdoor Attacks in Federated Learning 2
Cross Model Adversarial Transferability 2
Ensemble Strategies for Robust Membership Inference 1
Structured Textual Representations for GUI Automation 1
Cross Implementation Consistency Checking 1
Cross Layer State Desynchronization in Path MTU Discovery 1
Pixel-Based Grounding for Metadata-Independent Interface Navigation 1

Top Authors

AuthorPapers
Ke Xu5
Qi Li5
Shuai Wang5
Daoyuan Wu3
Tianwei Zhang3
Xiao Wang3
Zhuotao Liu3
Alina Oprea2
Bo Li2
D Weber2
Ee-Chien Chang2
Fabian Thomas2
Gefei Tan2
Jason Xue2
Jiahai Yang2
Jie Zhang2
Liangcai Su2
Mario Fritz2
Matthew Jagielski2
Michael Backes2

Impactful Papers

Papers from CCS 2025 that shaped research clusters.

Influential

Papers

Title
Prompt Inference Attack on Distributed Large Language Model Inference Frameworks
Prototype Surgery: Tailoring Neural Prototypes via Soft Labels for Efficient Machine Unlearning
Provable Repair of Deep Neural Network Defects by Preimage Synthesis and Property Refinement
Pseudocapacitive Enrichment of Interfacial Lithiophilicity for Anode-Free Lithium-Hydrogen Gas Battery
QRSec 2025: ACM CCS First Workshop on Quantum-Resistant Cryptography and Security
RAG-WM: An Efficient Black-Box Watermarking Approach for Retrieval-Augmented Generation of Large Language Models
Realizing Corrupted-Shard Tolerance: A Sharding Blockchain with Preserving Global Resilience
RebirthDay Attack: Reviving DNS Cache Poisoning with the Birthday Paradox
Refined TFHE Leveled Homomorphic Evaluation and Its Application
Removal Attack and Defense on AI-generated Content Latent-based Watermarking
Rethinking Machine Unlearning in Image Generation Models
Riddle Me This! Stealthy Membership Inference for Retrieval-Augmented Generation
Right the Ship: Assessing the Legitimacy of Invalid Routes in RPKI
RingSG: Optimal Secure Vertex-Centric Computation for Collaborative Graph Processing
RISCover: Automatic Discovery of User-exploitable Architectural Security Vulnerabilities in Closed-Source RISC-V CPUs
RVISmith: Fuzzing Compilers for RVV Intrinsics
Safeguarding Graph Neural Networks against Topology Inference Attacks
SafeGuider: Robust and Practical Content Safety Control for Text-to-Image Models
Scalable Cryptography for Trustworthy Machine Learning in the LLM Era
ScannerGrouper: A Generalizable and Effective Scanning Organization Identification System Toward the Open World
SecAlign: Defending Against Prompt Injection with Preference Optimization
Secure Noise Sampling for Differentially Private Collaborative Learning
Securing Mixed Rust with Hardware Capabilities
Security-Aware Sensor Fusion with MATE: the Multi-Agent Trust Estimator
Single-Cell Microgel Microrobot for Targeted and Imaging-Guided Cell Therapy
SISTAR: An Efficient DDoS Detection and Mitigation Framework Utilizing Programmable Data Planes
Slot: Provenance-Driven APT Detection through Graph Reinforcement Learning
Split Unlearning
Styled to Steal: The Overlooked Attack Surface in Email Clients
Supramolecular Chemistry for High-Performance Lithium Metal Batteries
Swallow: A Transfer-Robust Website Fingerprinting Attack via Consistent Feature Learning
Sylva: Tailoring Personalized Adversarial Defense in Pre-trained Models via Collaborative Fine-tuning
Systematic Assessment of Tabular Data Synthesis
TEMPEST-LoRa: Cross-Technology Covert Communication
TensorShield: Safeguarding On-Device Inference by Shielding Critical DNN Tensors with TEE
The 20th Workshop on Programming Languages and Analysis for Security (PLAS 2025)
The Importance of Being Discrete: Measuring the Impact of Discretization in End-to-End Differentially Private Synthetic Data
The Odyssey of robots.txt Governance: Measuring Convention Implications of Web Bots in Large Language Model Services
The Phantom Menace in Crypto-Based PET-Hardened Deep Learning Models: Invisible Configuration-Induced Attacks
THOR: Secure Transformer Inference with Homomorphic Encryption
Threat from Windshield: Vehicle Windows as Involuntary Attack Sources on Automotive Voice Assistants
Tide: An Efficient Kernel-level Isolation Execution Environment on AArch64 via Dynamically Adjusting Output Address Size
Time Tells All: Deanonymization of Blockchain RPC Users with Zero Transaction Fee
Towards Backdoor Stealthiness in Model Parameter Space
Towards Practical High Energy Density Lithium-Sulfur Batteries
Training Robust Classifiers for Classifying Encrypted Traffic under Dynamic Network Conditions
Training with Only 1.0 ‰ Samples: Malicious Traffic Detection via Cross-Modality Feature Fusion
Turning Uncertainty into Efficiency: Toward Practical, Quantum-Resistant Verifiable Privacy Tools
Unmask Tampering: Efficient Document Tampering Localization under Recapturing Attacks with Real Distortion Knowledge
UnsafeBench: Benchmarking Image Safety Classifiers on Real-World and AI-Generated Images

« prev Page 3/4 next »