TH
T. Hristov
info
Please Note
<p>This page displays the records of the person named above and is not linked to a unique person identifier. This record may need to be merged to a profile.</p>
2 records found
1
Cybersecurity attacks are increasingly sophisticated, while traditional, rule-based intrusion detectionsystems (IDS) remain prone to high false alert rates. This research explores temporal graph learning fornetwork intrusion detection, introducing a framework that combines temporal graph construction withGraph Attention Networks and recurrent modeling (GATv2 + LSTM). We evaluate on the LANL authentication logs and Zeek logs from the University of West Florida (UWF).
On LANL, our models (Try1/Try2) outperform state-of-the-art baselines for temporal link prediction,achieving high precision and robustness: Accuracy ≈ 0.994, F1 ≈ 0.993, AUC ≈ 0.993–0.998, AP ≈ 0.999.On Zeek data, edge prediction is sensitive to how malicious activity is distributed over time: a simple“Day" shuffling that preserves the temporal structure while also spreading the clusters of attack activity,yields large gains (e.g., Accuracy ≈ 0.969, AUC ≈ 0.996, F1 ≈ 0.959, AP ≈ 0.995), whereas random shufflingharms temporal dependencies and performance.
Extending to edge classification (benign vs. malicious) reveals a key limitation: despite high accuracy,AUC and AP remain low due to a tendency to label nearly all edges as benign under class imbalance andtemporal clustering, producing many false negatives. We test mitigation strategies (dropout, alternativeloss formulations with confidence weighting), which provide a small increase in stability but do notfully resolve the issue.
With our results, we find that the proposed temporal graph method is a strong fit for anomaly detectionvia edge prediction: robust across datasets, resilient to imbalance, and practically applicable. In contrast,edge classification currently lacks reliability for production without improved data balancing, graphconstruction, and training. ...
On LANL, our models (Try1/Try2) outperform state-of-the-art baselines for temporal link prediction,achieving high precision and robustness: Accuracy ≈ 0.994, F1 ≈ 0.993, AUC ≈ 0.993–0.998, AP ≈ 0.999.On Zeek data, edge prediction is sensitive to how malicious activity is distributed over time: a simple“Day" shuffling that preserves the temporal structure while also spreading the clusters of attack activity,yields large gains (e.g., Accuracy ≈ 0.969, AUC ≈ 0.996, F1 ≈ 0.959, AP ≈ 0.995), whereas random shufflingharms temporal dependencies and performance.
Extending to edge classification (benign vs. malicious) reveals a key limitation: despite high accuracy,AUC and AP remain low due to a tendency to label nearly all edges as benign under class imbalance andtemporal clustering, producing many false negatives. We test mitigation strategies (dropout, alternativeloss formulations with confidence weighting), which provide a small increase in stability but do notfully resolve the issue.
With our results, we find that the proposed temporal graph method is a strong fit for anomaly detectionvia edge prediction: robust across datasets, resilient to imbalance, and practically applicable. In contrast,edge classification currently lacks reliability for production without improved data balancing, graphconstruction, and training. ...
Cybersecurity attacks are increasingly sophisticated, while traditional, rule-based intrusion detectionsystems (IDS) remain prone to high false alert rates. This research explores temporal graph learning fornetwork intrusion detection, introducing a framework that combines temporal graph construction withGraph Attention Networks and recurrent modeling (GATv2 + LSTM). We evaluate on the LANL authentication logs and Zeek logs from the University of West Florida (UWF).
On LANL, our models (Try1/Try2) outperform state-of-the-art baselines for temporal link prediction,achieving high precision and robustness: Accuracy ≈ 0.994, F1 ≈ 0.993, AUC ≈ 0.993–0.998, AP ≈ 0.999.On Zeek data, edge prediction is sensitive to how malicious activity is distributed over time: a simple“Day" shuffling that preserves the temporal structure while also spreading the clusters of attack activity,yields large gains (e.g., Accuracy ≈ 0.969, AUC ≈ 0.996, F1 ≈ 0.959, AP ≈ 0.995), whereas random shufflingharms temporal dependencies and performance.
Extending to edge classification (benign vs. malicious) reveals a key limitation: despite high accuracy,AUC and AP remain low due to a tendency to label nearly all edges as benign under class imbalance andtemporal clustering, producing many false negatives. We test mitigation strategies (dropout, alternativeloss formulations with confidence weighting), which provide a small increase in stability but do notfully resolve the issue.
With our results, we find that the proposed temporal graph method is a strong fit for anomaly detectionvia edge prediction: robust across datasets, resilient to imbalance, and practically applicable. In contrast,edge classification currently lacks reliability for production without improved data balancing, graphconstruction, and training.
On LANL, our models (Try1/Try2) outperform state-of-the-art baselines for temporal link prediction,achieving high precision and robustness: Accuracy ≈ 0.994, F1 ≈ 0.993, AUC ≈ 0.993–0.998, AP ≈ 0.999.On Zeek data, edge prediction is sensitive to how malicious activity is distributed over time: a simple“Day" shuffling that preserves the temporal structure while also spreading the clusters of attack activity,yields large gains (e.g., Accuracy ≈ 0.969, AUC ≈ 0.996, F1 ≈ 0.959, AP ≈ 0.995), whereas random shufflingharms temporal dependencies and performance.
Extending to edge classification (benign vs. malicious) reveals a key limitation: despite high accuracy,AUC and AP remain low due to a tendency to label nearly all edges as benign under class imbalance andtemporal clustering, producing many false negatives. We test mitigation strategies (dropout, alternativeloss formulations with confidence weighting), which provide a small increase in stability but do notfully resolve the issue.
With our results, we find that the proposed temporal graph method is a strong fit for anomaly detectionvia edge prediction: robust across datasets, resilient to imbalance, and practically applicable. In contrast,edge classification currently lacks reliability for production without improved data balancing, graphconstruction, and training.
Although digital watermarking has been a well-researched topic for the past decades and has seen numerous implementations for relational databases, it still lacks research for non-relational schema-less databases. In this paper, we explore proposed techniques for non-relational database watermarking and introduce an improved technique for NoSQL database watermarking. The improved technique produces pseudo documents, embeds a watermark in them, and inserts them into the database. Then it modifies the relations between the pseudo and real documents to indicate which documents are covered by the watermark. Lastly, the pseudo documents act as a cover for the watermark embedding, proving the authenticity of the documents connected. We tested the technique against numerous modification and deletion attacks. The results show that the introduced technique does not improve the robustness of the watermark. Although the technique has the potential to produce a robust watermark, it still needs to be improved before being used in a commercial setting.
...
Although digital watermarking has been a well-researched topic for the past decades and has seen numerous implementations for relational databases, it still lacks research for non-relational schema-less databases. In this paper, we explore proposed techniques for non-relational database watermarking and introduce an improved technique for NoSQL database watermarking. The improved technique produces pseudo documents, embeds a watermark in them, and inserts them into the database. Then it modifies the relations between the pseudo and real documents to indicate which documents are covered by the watermark. Lastly, the pseudo documents act as a cover for the watermark embedding, proving the authenticity of the documents connected. We tested the technique against numerous modification and deletion attacks. The results show that the introduced technique does not improve the robustness of the watermark. Although the technique has the potential to produce a robust watermark, it still needs to be improved before being used in a commercial setting.