Security in AI Comparison Hub
6 papers - avg viability 4.2
Top Papers
- BackdoorIDS: Zero-shot Backdoor Detection for Pretrained Vision Encoder(8.0)
BackdoorIDS offers a zero-shot method for detecting backdoor attacks in pretrained vision encoders, enhancing security in AI applications.
- LoRA as Oracle(5.0)
A LoRA-based framework for detecting backdoor and privacy threats in neural networks for security-critical applications.
- Delayed Backdoor Attacks: Exploring the Temporal Dimension as a New Attack Surface in Pre-Trained Models(5.0)
Introducing Delayed Backdoor Attacks, a novel threat model for pre-trained models that exploits temporal dimensions for malicious activation.
- KEPo: Knowledge Evolution Poison on Graph-based Retrieval-Augmented Generation(4.0)
KEPo is a novel poisoning attack method designed to exploit vulnerabilities in Graph-based Retrieval-Augmented Generation systems.
- Detecting and Eliminating Neural Network Backdoors Through Active Paths with Application to Intrusion Detection(3.0)
A novel approach to detect and eliminate backdoor triggers in machine learning models for intrusion detection.
- Security Considerations for Multi-agent Systems(2.0)
This study evaluates security frameworks for multi-agent AI systems, highlighting vulnerabilities and gaps in existing solutions.