Large Language Model (LLM) security

Large Language Models (LLMs) have become the foundation of modern Artificial Intelligence, enabling breakthroughs in natural language processing, intelligent assistants, code generation, decision support, cybersecurity automation, and scientific research. As LLMs are increasingly deployed in critical domains such as healthcare, finance, government, defense, education, and cloud services, ensuring their security, privacy, robustness, and trustworthiness has become a major research challenge. These models are vulnerable to a wide range of security threats, including prompt injection and jailbreak attacks, adversarial inputs, model poisoning, data leakage, model extraction, membership inference, hallucinations, malicious fine-tuning, and supply chain attacks. Addressing these challenges is essential for the safe and reliable deployment of LLM-powered applications.

The International Conference on Cyber Security in Artificial Intelligence (ICCSAI 2027) invites original research contributions on the security of Large Language Models (LLMs). This track aims to bring together researchers, cybersecurity professionals, AI developers, industry practitioners, and policymakers to discuss innovative approaches for protecting LLMs throughout their lifecycle—from data collection and model training to deployment, monitoring, and governance. Contributions presenting novel attack methodologies, defensive mechanisms, secure architectures, privacy-preserving techniques, evaluation frameworks, governance models, and real-world applications are highly encouraged.

1. LLM Architecture and Foundation Model Security

• Security of Large Language Models (LLMs)

• Foundation Model Security

• Secure LLM Architectures

• Model Alignment and AI Safety

• Robustness of Foundation Models

2. Prompt and Interaction Security

• Prompt Injection Attacks

• Jailbreak Attacks and Defenses

• System Prompt Leakage

• Prompt Engineering Security

• Context Window Manipulation

• Indirect Prompt Injection

• Prompt Filtering and Validation

3. Adversarial Attacks on LLMs

• Adversarial Prompting

• Evasion Attacks

• Backdoor Attacks

• Data Poisoning and Training-Time Attacks

• Adversarial Fine-Tuning

• Robustness Against Adversarial Inputs

4. Model Privacy and Data Protection

• Membership Inference Attacks

• Model Inversion Attacks

• Sensitive Information Disclosure

• Privacy-Preserving LLM Training

• Differential Privacy for LLMs

• Confidential AI and Secure Inference

5. Model Theft and Intellectual Property

• Model Extraction Attacks

• Model Stealing and Cloning

• Intellectual Property Protection

• Model Watermarking and Fingerprinting

• AI Model Licensing and Protection

6. Secure Fine-Tuning and Deployment

• Secure Fine-Tuning Techniques

• Parameter-Efficient Fine-Tuning (PEFT) Security

• Reinforcement Learning from Human Feedback (RLHF) Security

• Secure Model Serving

• LLM Deployment Security

• Secure AI APIs

7. Retrieval-Augmented Generation (RAG) Security

• Secure RAG Architectures

• Vector Database Security

• Retrieval Poisoning Attacks

• Knowledge Base Integrity

• Secure Retrieval Pipelines

8. AI Agent and Tool Security

• LLM-Based Autonomous Agents

• Tool Calling Security

• Agent Authentication and Authorization

• Multi-Agent System Security

• Secure Human–AI Collaboration

9. AI-Enabled Cyber Threats

• AI-Generated Malware

• AI-Assisted Phishing

• Social Engineering Using LLMs

• Detection of Malicious AI Content

• Deepfake Text and Synthetic Content Detection

10. Trustworthy and Explainable LLMs

• Hallucination Detection and Mitigation

• Explainable LLMs

• AI Transparency and Accountability

• Trust Calibration

• Bias, Fairness, and Security

11. Secure AI Operations

• Secure MLOps and LLMOps

• AI DevSecOps

• Continuous Security Monitoring

• AI Supply Chain Security

• Third-Party Model Risk Management

12. Governance, Compliance, and Risk

• AI Governance Frameworks

• LLM Risk Assessment

• Regulatory Compliance

• AI Security Standards

• Ethical and Responsible LLM Deployment

• Security Auditing and Certification

13. Emerging Topics in LLM Security

• Multimodal LLM Security

• Edge LLM Security

• Federated LLM Security

• Quantum-Resistant AI Security

• Zero-Trust Architectures for LLMs

• AI Security Red Teaming

• LLM Security Benchmarks and Evaluation

• Secure Open-Source LLM Ecosystems

• AI Content Provenance and Watermarking

• Resilient and Self-Healing AI Systems