Adversarial attacks against AI models

Adversarial Attacks Against AI Models

Artificial Intelligence (AI) has become a cornerstone of modern cybersecurity, autonomous systems, healthcare, finance, and critical infrastructure. However, the growing adoption of AI has also introduced new security challenges, particularly adversarial attacks that manipulate machine learning models through carefully crafted inputs, data poisoning, model evasion, and inference attacks. These attacks can significantly degrade model performance, compromise decision-making, and threaten the reliability of AI-powered applications. As AI systems are increasingly deployed in security-critical environments, developing robust, trustworthy, and resilient AI models capable of withstanding adversarial manipulation has become a major research priority.

The International Conference on Cyber Security in Artificial Intelligence (ICCSAI 2027) invites researchers, academicians, cybersecurity professionals, AI practitioners, industry experts, and policymakers to present innovative research, methodologies, and practical solutions addressing adversarial attacks against AI models. This track focuses on adversarial machine learning, adversarial example generation, model poisoning, evasion attacks, backdoor attacks, model extraction, model inversion, membership inference, privacy attacks, secure federated learning, explainable AI (XAI), trustworthy AI, robust deep learning, AI model verification, and defense mechanisms for machine learning systems. The track also explores the security of large language models (LLMs), generative AI, autonomous systems, computer vision, natural language processing, and AI-enabled cyber defense.

Participants will discuss emerging threats, novel attack techniques, resilient AI architectures, secure model deployment strategies, and regulatory frameworks for trustworthy AI. Special emphasis is placed on securing AI applications across cloud computing, edge intelligence, Internet of Things (IoT), critical infrastructure, healthcare, finance, smart cities, and autonomous transportation. Through keynote lectures, technical sessions, and interdisciplinary collaboration, ICCSAI 2027 provides a premier platform for advancing research that strengthens the security, privacy, robustness, and resilience of AI systems against adversarial threats, ultimately enabling safer and more dependable intelligent technologies for the future.

1. Adversarial Machine Learning
2. Adversarial Example Generation
3. Evasion Attacks
4. Data Poisoning Attacks
5. Model Poisoning in Federated Learning
6. Backdoor and Trojan Attacks
7. Model Extraction Attacks
8. Model Inversion Attacks
9. Membership Inference Attacks
10. Privacy Attacks on AI Models
11. Adversarial Training
12. Defensive Distillation
13. Certified Robustness Methods
14. Explainable AI for Robust AI
15. Trustworthy AI Systems
16. AI Model Verification and Validation
17. Robust Deep Learning Architectures
18. Security of Large Language Models (LLMs)
19. Prompt Injection and Jailbreak Attacks
20. Security of Generative AI Models
21. AI Supply Chain Security
22. Secure MLOps Pipelines
23. AI Model Monitoring and Drift Detection
24. AI Threat Intelligence
25. Benchmarking Adversarial Robustness
26. Privacy-Preserving Machine Learning
27. Secure Federated Learning
28. AI Risk Assessment
29. AI Security Standards and Compliance
30. Emerging Challenges in Adversarial AI