AAAI 2026 accepted papers on Adversarial robustness and security
137 of the 4,417 papers accepted at AAAI 2026 (3.1%) are on Adversarial robustness and security, compared with 3.0% at AAAI 2025. They fall into 4 topics.
See AAAI 2026 in the interactive map
Working on something in this area? Describe your idea in scime atlas to see the closest accepted papers by meaning and how crowded the spot has become.
Topics and papers
Attacks on text and graphs · 76 papers
- Uncovering and Mitigating Destructive Multi-Embedding Attacks in Deepfake Proactive Forensics (3 citations)
- ARIW-Framework: Adaptive Robust Iterative Watermarking Framework (2 citations)
- Dual-View Inference Attack: Machine Unlearning Amplifies Privacy Exposure (2 citations)
- CluCERT: Certifying LLM Robustness via Clustering-Guided Denoising Smoothing (1 citation)
- Cost-Minimized Label-Flipping Poisoning Attack to LLM Alignment (1 citation)
- CP-FREEZER: Latency Attacks Against Vehicular Cooperative Perception (1 citation)
- Creating Blank Canvas Against AI-enabled Image Forgery (1 citation)
- DeepTracer: Tracing Stolen Model via Deep Coupled Watermarks (1 citation)
- Disentangling Adversarial Prompts: A Semantic-Graph Defense for Robust LLM Security (1 citation)
- Edge Self-Adversarial Augmentation Enhances Graph Contrastive Learning Against Neighborhood Inconsistency (1 citation)
- GeoShield: Safeguarding Geolocation Privacy from Vision-Language Models via Adversarial Perturbations (1 citation)
- IO-RAE: Information-Obfuscation Reversible Adversarial Example for Audio Privacy Protection (1 citation)
and 64 more in the interactive map.
Backdoor and poisoning · 28 papers
- Clean-Label Physical Backdoor Attacks with Data Distillation (2 citations)
- ConfGuard: A Simple and Effective Backdoor Detection for Large Language Models (2 citations)
- Towards Effective, Stealthy, and Persistent Backdoor Attacks Targeting Graph Foundation Models (2 citations)
- Backdoor Attacks on Open Vocabulary Object Detectors via Multi-Modal Prompt Tuning (1 citation)
- DIFT: Protecting Contrastive Learning Against Data Poisoning Backdoor Attacks (1 citation)
- DUP: Detection-guided Unlearning for Backdoor Purification in Language Models (1 citation)
- Enhancing All-to-X Backdoor Attacks with Optimized Target Class Mapping (1 citation)
- Hidden in the Noise: Unveiling Backdoors in Audio LLMs Alignment Through Latent Acoustic Pattern Triggers (1 citation)
- Modulation-Based Backdoors: Leveraging Amplitude and Frequency Patterns to Attack Speaker Recognition (1 citation)
- PurMM: Attention-Guided Test-Time Backdoor Purification in Multimodal Large Language Models (1 citation)
- Transferable Backdoor Attacks for Code Models via Sharpness-Aware Adversarial Perturbation (1 citation)
- Uncovering and Aligning Anomalous Attention Heads to Defend Against NLP Backdoor Attacks (1 citation)
and 16 more in the interactive map.
Certified and empirical robustness · 17 papers
- MPD-SGR: Robust Spiking Neural Networks with Membrane Potential Distribution-Driven Surrogate Gradient Regularization (1 citation)
- On Robustness of Linear Classifiers to Targeted Data Poisoning (1 citation)
- A Flat Minima Perspective on Understanding Augmentations and Model Robustness
- ALERT: Adversarial Learning Enhanced Stability-aware Routing Transformer for Adaptive Depression Detection
- Angular Gradient Sign Method: Uncovering Vulnerabilities in Hyperbolic Networks
- Boosting the Robustness-Accuracy Trade-off of SNNs by Robust Temporal Self-Ensemble
- Certified but Fooled! Breaking Certified Defenses with Ghost Certificates
- CertMask: Certifiable Defense Against Adversarial Patches via Theoretically Optimal Mask Coverage
- Debiased Dual-Invariant Defense for Adversarially Robust Person Re-Identification
- Diversifying Counterattacks: Orthogonal Exploration for Robust CLlP Inference
- DRFGD: Disentangled Representation-Focused Generative Defense for Attack-Tolerant Cross-Modal Hashing
- Exploiting Missing Data Remediation Strategies Using Adversarial Missingness Attacks
and 5 more in the interactive map.
Black-box and transfer attacks · 16 papers
- Boosting Adversarial Transferability via Ensemble Non-Attention (1 citation)
- Fading the Digital Ink: A Universal Black-Box Attack Framework for 3DGS Watermarking Systems (1 citation)
- Transferability of Adversarial Attacks in Video-based MLLMs: A Cross-modal Image-to-Video Approach (1 citation)
- Transferable Hypergraph Attack via Injecting Nodes into Pivotal Hyperedges (1 citation)
- Adversarial Attack on Black-Box Multi-Agent by Adaptive Perturbation
- An Information Theoretic Evaluation Metric for Strong Unlearning
- From Pretrain to Pain: Adversarial Vulnerability of Video Foundation Models Without Task Knowledge
- HogVul: Black-box Adversarial Code Generation Framework Against LM-based Vulnerability Detectors
- Improving the Convergence Rate of Ray Search Optimization for Query-Efficient Hard-Label Attacks
- Prompting Adversarial Transferability via Path Flatness Attack
- Removing Box-Free Watermarks for Image-to-Image Models via Query-Based Reverse Engineering
- STACK: Adversarial Attacks on LLM Safeguard Pipelines
and 4 more in the interactive map.