Call for Papers: Advancements in Privacy, Security, and Integrity of Neural Networks: Attacks and Defences

The widespread deployment of deep neural networks (DNNs) in healthcare, finance, autonomous systems, and generative AI has heightened concerns about security, robustness, and integrity. AI models are increasingly vulnerable to adversarial perturbations, poisoning, backdoor attacks, unauthorized retraining, and model extraction, which can silently compromise performance in safety-critical applications.

Since trained models represent significant intellectual and computational investments, protecting ownership, detecting tampering, and verifying authenticity have become key research priorities. Techniques such as watermarking, fingerprinting, reversible data hiding, and forensic analysis offer promising solutions without degrading model performance.

The rapid growth of generative AI further raises issues of authenticity and provenance. Integrity-preserving mechanisms, along with complementary tools such as blockchain-based audit trails, can support secure verification and accountability.

List of Topic Areas

This Special Issue invites contributions on adversarial robustness, integrity-preserving methods, and secure verification frameworks for neural networks and AI-generated systems. Topics of interest include, but are not limited to:

Adversarial perturbation detection and certified defences
Model poisoning and backdoor attack mitigation
Integrity verification and tamper detection in neural networks
Digital watermarking and fingerprinting for model protection
Neural network forensics and authenticity verification
Detection of unauthorized retraining and model extraction
Security of generative AI systems
Privacy-preserving and secure neural network design
Blockchain-supported model provenance and audit mechanisms (complementary)

Submission and Review Process

All submitted manuscripts will undergo a formal single-blind peer-review process. Papers will be handled on a first-come, first-served basis. Accepted papers will be published open access upon acceptance and later compiled into the Special Issue collection. Manuscripts not accepted within the publication window may be transferred to the journal’s regular track.

To submit your paper, please see here