Quantum classical link prediction tested on dark web link graphs
Quantum-Classical Link Prediction on Dark Web Hyperlink Graphs
Networking and Internet ArchitectureMachine Learning
Summary
Predicting connections between websites on the dark web is hard because of the hidden and complex structure. The authors studied a small network of websites and tried different ways to guess links, including classical machine learning and a quantum classifier. Classical graph-based models did well, but the quantum classifier performed similarly to simple classical methods on this data. Their work shows how tricky it is to compare quantum and classical approaches fairly, especially when dealing with real-world networks and content features.
What this means in practice
- •For cybersecurity analysts: Use combined graph and content features to identify potential hidden links between dark web hosts for threat intelligence.
- •For machine learning engineers: Improve link prediction models by incorporating degree-controlled negative sampling and multimodal data from network graphs and text content.
Tested on one dataset.
Authors
Vojtěch Novák, Radovan Rečka, Ondřej Knebl, Aktham Youssef, Ivan Zelinka
Abstract
We analyze HTML content and hyperlinks from an offline Tor crawl containing 383 onion hosts and 2,067 directed cross-host links. The study combines semantic analysis with degree-controlled quantum--classical link prediction. Two directory groups emit 94.8\% of the observed host links. Directed preferential attachment reaches ROC-AUC 0.976 with random non-edges, but falls to $0.550\pm0.014$ under source-conditioned target-degree matching. With disjoint training and test pairs, multimodal logistic regression reaches $0.672\pm0.021$, GAE $0.899\pm0.016$, and VGAE $0.906\pm0.016$. A four-qubit variational quantum classifier (VQC) comparison tests two encodings across 20 disjoint-pair graph splits under exact and 1,024-shot simulation, with classical controls using the same inputs and training examples. Its best exact mean ROC-AUC is $0.549\pm0.023$, compared with $0.551\pm0.016$ for RBF SVC and $0.573\pm0.029$ for a 16-parameter MLP on the same inputs and training examples. These results support predictive signal in combined graph/content features and show why degree control, feature compression, and optimization conditions matter in quantum--classical comparisons.