Prompt Injection Attack on LLM-integrated Applications: Vulnerabilities Unveiled
Prompt Injection Attacks pose significant security risks to LLM-integrated applications, with potential severe outcomes.
LLM-integrated Applications Vulnerable to Prompt Injection Attacks
Prompt injection attacks pose significant security risks to LLM-integrated applications, highlighting vulnerabilities and potential exploitation.
Optimal Sensor Deception for Itinerary Deviation Study
Deceptive sensor alteration is crucial for disguising adversarial itineraries.
Pre-processing Defenses for White-box Robustness Enhancement
Deep neural networks are vulnerable to adversarial noise, and pre-processing methods can enhance white-box robustness by utilizing full adversarial examples.
SoK: Essential Guide for Malware Sandbox Usage in Security Applications
Malware sandboxes are essential for security applications, but their complexity can impact results significantly. Systematizing sandbox practices and guidelines can improve the effectiveness of sandbox deployments.
EasyJailbreak: A Unified Framework for Jailbreaking Large Language Models
EasyJailbreak introduces a modular framework simplifying jailbreak attacks against Large Language Models, revealing vulnerabilities and emphasizing the need for enhanced security measures.
Imperio: Language-Guided Backdoor Attacks for Arbitrary Model Control
Natural language understanding enhances backdoor attacks in NLP models, as demonstrated by Imperio.
Skeptic: Automatic, Justified and Privacy-Preserving Password Composition Policy Selection
提案された方法論に基づいて、パスワード組成ポリシーを自動的に選択し、正当化し、プライバシーを保護する。
Physical Layer Supply Voltage Coupling Vulnerability for Information Leakage
PSVC vulnerability allows data extraction without device modification, demonstrated through end-to-end attacks on microcontrollers.
GNN-based Control-Flow Attestation for Embedded Devices: Overcoming Limitations with RAGE
RAGE introduces a novel, lightweight CFA approach for embedded devices, addressing limitations of existing schemes.
© 2024 by Linnk AI