ML4D – Machine Learning for Defence

About the Conference

Artificial intelligence is moving from research prototypes into systems that matter for security and defence — from autonomous platforms and multi-sensor situational awareness to logistics, cyber defence, and decision support. These applications raise challenges that general-purpose AI research rarely addresses: adversaries that deliberately attack the model, data that is scarce and sensitive, assurance requirements that must hold under operational conditions, and platforms that must run at the edge with limited power and unreliable communication. ML4D aims to examine these requirements and explore potential AI-driven approaches for security and defence. Key topics include AI use cases in security and defense, innovative methodologies for addressing sector-specific challenges, and both technical and non-technical challenges associated with AI implementation in these contexts. The workshop brings together researchers from universities, research institutes, industry, and the armed forces to discuss recent advances, open problems, and future directions.

Register here: TBA

The first Machine Learning for Defence (ML4D) workshop will take place on March 12, 2027 at the Fraunhofer Forum in Berlin. ML4D will be held jointly with the 11th Machine Learning for Cyber-Physical Systems (ML4CPS) workshop on March 11, 2027.

ML4D is hosted by Fraunhofer IOSB and FKIE, Helmut Schmidt University Hamburg, the German Aerospace Center (DLR), and the University of the Bundeswehr Munich.

Papers may cover, but are not limited to, the following topics:

• Autonomy & Agentic AI for Defence: Autonomous, tool-using agents that independently plan, diagnose, and act reach beyond text- and multimodal-focused LLM agents, raising questions of mission-level autonomy, human-machine teaming, and meaningful human control.
• Symbolic and Neuro-Symbolic AI: Knowledge representation, automated planning, constraint reasoning, and model-based diagnosis, together with their combination with learned components.
• Multi-Sensor Fusion and Situational Awareness: Detection, classification, and tracking from heterogeneous sources such as EO/IR, radar, acoustic, and textual data, under real-time constraints and with incomplete or contradictory observations.
• Assurance, Verification, and Certification: Test and evaluation, formal verification, uncertainty quantification, and explainability for systems that must be certified and must comply with legal and ethical requirements, including international humanitarian law.
• Learning from Scarce, Sensitive, and Synthetic Data: Operational data is rare, restricted, or classified. Simulation-to-real transfer, synthetic data generation, and federated or privacy-preserving learning across security domains are key enablers.
• Predictive Maintenance, Logistics, and Readiness: AI methods for the availability and sustainment of military platforms and fleets, connecting the defence domain to established cyber-physical systems methods.
• Edge AI under Size, Weight, and Power Constraints: On-platform inference within tight compute and energy budgets, and reliable operation in degraded, disconnected, or low-bandwidth environments.
• AI for Cyber Defence: Anomaly and intrusion detection in military networks and command-and-control infrastructure, including the protection of the AI components themselves.

Agenda

TBA

Conference Location

Fraunhofer Forum Berlin

Anna-Louisa-Karsch-Straße 2

10178 Berlin

Hosts

Important Dates

Title & Abstract Registration: December 11th, 2026

Full Paper Submission: December 18th, 2026

Notification of Acceptance: January 18th, 2027

Camera-Ready Paper: February 26th, 2027

Submission Guidelines

All papers undergo a single-blind peer-review process. Authors first register the title and abstract of their contribution through the submission portal and submit the full paper (max. 15 pages) one week later. Accepted papers are invited to be published in the workshop proceedings by Helmut Schmidt University Press (openHSU) and receive a unique DOI. All submissions must be unclassified and cleared for public release by the authors’ organisations. Obtaining this clearance is the responsibility of the authors.

Please use the following template for your submission:

TBA

Paper Submission will be handled via easychair:

TBA

For additional details and submission guidelines, please refer to

[email protected]

Committee

TBA