Evidence record · UDINUS · SDGs
- Unit
- Rektorat UDINUS
- Period
- 2026
- Goals
- Record
- 54a867b2-7286-4d05-8028-5c23f21082c0
- Published
- 2026-10-30
Universitas Dian Nuswantoro (UDINUS), through researchers Sahu, Aditya Kumar; Dhar, Swapnaneel; Deb, Subhrajyoti; Khaldi, Amine; Sahu, Monalisa; and Setiadi, De Rosal Ignatius Moses, conducted an activity on AI-assisted, simulation-based quantum image watermarking using Novel Enhanced Quantum Representation (NEQR) and reversible quantum logic. The activity addressed limitations associated with Flexible Representation of Quantum Images (FRQI), including high computational and qubit complexity, as well as limitations in scalability. It also considered vulnerabilities of quantum comparator models to geometric attacks such as scaling and rotation and their slow embedding and extraction rate.
The proposed technique integrated Quantum Image Representation with NEQR. Watermark qubit generation combined pixel position bits with an arithmetic average-based logic through a quantum circuit comprising reverse parallel adder, dividing-4, and controlled-NOT models. Watermark embedding used reverse parallel adder, quantum equal models, controlled-NOT gates, and Pauli-X gates, applying a least significant bit strategy within an NEQR-encoded image. Extraction mirrored the embedding architecture and used a reverse parallel subtractor, quantum equal models, controlled-NOT gates, and Pauli-X gates.
The activity was validated through IBM Qiskit-based quantum circuit simulation on classical computing hardware. The source explicitly states that it was not intended to demonstrate fault-tolerant quantum hardware execution or provable quantum computational advantage. Experimental results reported a highest peak-signal-to-noise ratio of 59.97 dB and a structural similarity index matrix of 0.9999. The technique was evaluated against image processing attacks, quantum noise, and decoherence. An AI-assisted attack classification module using a lightweight Random Forest classifier identified categories including no attack, Gaussian noise, salt-and-pepper noise, median filtering, histogram equalization, JPEG compression, cropping, rotation, and text addition, with an overall classification accuracy of 96.56%.
In relation to the stated RAW_GOALS, the source documents a simulation-level research activity in secure image watermarking and attack identification. The activity contributes institutional evidence of UDINUS research using reversible quantum-circuit modelling and AI-assisted classification, while the source does not state a direct connection with a specific substantive target of the stated goal.