Dr Mahender Kumar

AI Safety & Trustworthy AI

AI assurance, reliability, human-AI teaming, model evaluation, explainability and evidence-grounded AI.

Reliability · Assurance · Human-AI Decision Making

Cybersecurity & Critical Systems

Cyber-physical security, automotive and CAN security, medical IoT, additive-manufacturing supply chains and threat intelligence.

Critical Infrastructure · Automotive · Healthcare · Manufacturing

Cryptography, Privacy & Secure Systems

The technical foundation beneath the work above: elliptic-curve and identity-based cryptography, authentication and key exchange, and privacy-preserving analytics.

Homomorphic Encryption · ZKP · Authentication · Secure Data Sharing

ATA-SOC-HAT (Computers & Security, 2026) and MAGD-Fraud examine when AI recommendations should be trusted, overridden or deferred to a human reviewer.

FinAgent-Reliance is a human-in-the-loop reliability benchmark for financial LLM agents performing multi-step reasoning.

ICSThreatQA grounds AI-generated answers in verifiable knowledge for industrial control system threat intelligence; UK NACE applies explainable AI to automotive intrusion prevention.

A developing research direction on how organisations should evaluate AI systems, and preserve evidence provenance, before deployment in high-stakes environments.

A Threat, Vulnerability and Risk Assessment (TVRA) toolbox for medical IoT, coordinated across 12+ European partners and aligned with ISO 27001 and EU medical-device regulation.

Research areas: Medical IoT · AI Assurance · Cybersecurity

A real-time risk-scoring and intrusion-prevention engine for connected vehicles, using explainable AI and Deep Reinforcement Learning. Delivered 98% detection accuracy with a 22% reduction in false positives, aligned with ISO/SAE 21434.

Research areas: Automotive Security · Explainable AI

Attack-path detection and blockchain-based dual-lock integrity auditing for additive-manufacturing supply chains, with privacy-preserving threat-intelligence sharing.

Research areas: Manufacturing Security · Blockchain · Cryptography

A Science & Technology Ontology of 5,153 emerging-technology topics, built with BERTopic and knowledge graphs for semi-automated horizon scanning.

Research areas: NLP · Knowledge Graphs