Science

I work at the boundary between medical and industrial imaging, uncertainty, and decision-making in secure way. 

I work at the intersection of cybersecurity and advanced data technologies, with a focus on cybersecurity for 3D image processing, AI systems, and database infrastructures. My research also includes 3D medical image processing, combining data integrity, security, and reliable analysis of complex imaging data.

Adversarial Attacks in Industry 4.0

Industrial production systems increasingly depend on deep learning–based machine-vision models for automated inspection, quality assurance, and process monitoring in Industry 4.0 environments. While these models achieve high accuracy under normal conditions, they are vulnerable to adversarial attacks that introduce subtle, visually imperceptible perturbations. Such attacks can lead to defect masking, misclassification, and potentially unsafe system behaviour. Despite extensive research on adversarial robustness in generic vision tasks, systematic studies tailored to industrial machine-vision pipelines remain limited.

Publication

3D Image processing: 

Three‑dimensional (3D) perception is widely used in robotics, autonomous vehicles and medicine. These systems often operate autonomously and therefore are vulnerable to data tampering or spoofing attacks. Existing work on 2D adversarial detection does not generalise well to the high‑dimensional and structured nature of 3D point clouds. Secure3D‑CV is a lightweight extension to OpenCV that monitors incoming 3D data streams, extracts robust features and computes risk scores to flag unsafe scenes.

Publication 

Cybersecurity perspective for 3D imaging:

1. Threat Surface at Image Creation
   The moment a 3D medical image is generated, it becomes a sensitive digital asset. The imaging device itself is a potential attack vector. Security begins with device hardening, firmware integrity, network segmentation, and strict authentication on the scanner.
   
   

2. Secure Network Architecture
   When 3D images are transmitted to PACS systems, hospital servers, or cloud platforms, encrypted communication protocols protect against interception and man-in-the-middle attacks. Network segmentation isolates imaging systems from general IT infrastructure to reduce lateral movement risk.
   
   

3. Encryption at Rest
   Once stored, 3D medical images are encrypted on servers or cloud storage. Key management systems control cryptographic keys separately from the data. This ensures that even if storage is compromised, the data remains unreadable.

4. Identity and Access Management
   Access to medical images is governed by strong authentication mechanisms such as multi-factor authentication and role-based access control. Least-privilege principles limit user access strictly to necessary datasets. Privileged access is monitored and restricted.
   
   

5. Data Minimization and Pseudonymization
   Where possible, identifying metadata is separated or pseudonymized. This reduces risk in case of unauthorized exposure. Cybersecurity controls ensure that personal identifiers are not unnecessarily replicated across systems.
   
   

6. Monitoring and Intrusion Detection
   Security systems continuously monitor access patterns, file transfers, and unusual behavior. Intrusion detection and anomaly detection systems flag suspicious activities such as bulk downloads or unusual login attempts.
   
   

7. Logging and Audit Trails
   Every access event is recorded. Audit trails ensure traceability, support compliance with regulations such as GDPR or HIPAA, and allow forensic analysis in case of incidents.
   
   

8. Backup and Ransomware Protection
   Secure, immutable backups protect against ransomware attacks. Backup systems are isolated and regularly tested to ensure recoverability of critical medical imaging data.
   
   

9. Governance and Compliance
   Cybersecurity policies define retention periods, access procedures, incident response protocols, and regulatory compliance requirements. Security is not only technical but also organizational.

Protecting DICOM Astrocytoma MRI Images from Cybersecurity Threats

A 3D DICOM MRI scan of an astrocytoma is highly sensitive medical data. It contains detailed brain imaging, patient identifiers, and clinical metadata. Protecting it is not only an IT issue but a patient safety requirement.

1. Secure the Source

Protection begins at the MRI scanner.
The imaging device must run a hardened operating system, receive regular security patches, and operate within a segmented medical network. Only authorized technicians should access the system.

2. Protect Data in Transit

When DICOM files are transferred to PACS or cloud storage, encryption must be enforced.
DICOM over TLS, certificate-based authentication, and secure internal hospital networks prevent interception or manipulation during transmission.

3. Encrypt Data at Rest

Stored astrocytoma images must be encrypted using strong standards such as AES-256.
Encryption keys should be managed separately. Secure and immutable backups ensure recoverability in case of ransomware attacks.

4. Control Access Strictly

Only radiologists and authorized clinicians should access tumor images.
Role-based access control, multi-factor authentication, and least-privilege policies reduce insider and credential-based risks.

5. Secure AI Processing

If AI is used for tumor segmentation or volumetric analysis, the processing environment must be isolated.
Data anonymization, secure APIs, containerization, and monitoring prevent unauthorized data leakage.

6. Continuous Monitoring

Security systems must monitor access patterns and detect anomalies such as bulk downloads or unusual login attempts.
Audit logs ensure traceability and compliance with GDPR or HIPAA.

7. Ransomware & Integrity Protection

Immutable backups, network segmentation, and endpoint detection systems protect availability.
Digital signatures and integrity checks ensure that tumor images are not altered.