Temporal Quantum Neural Networks for Insider Threat Detection

Cyber-security plays a major role in protecting sensitive information in this digital era. In Cyber attacks identifying the internal threats are extremely difficult than the external one as the attack is performed the people who is having access to the sensitive information of the organization. Moreover due to diversity & volume of data, motive of the insider and rarity of attack makes it very hard to identify such insiders. The existing insider threat detection approaches has limitations like lower precision and granularity of features to be considered. In this regard, Principal Component Analysis(PCA) based dimensionality reduction and Quantum Neural Network(QNN) based detection approach has been proposed. The SEI CMU’s benchmark CERTr4.2 dataset is considered for evaluating this proposed approach. The feature vectors are constructed based on the day-wise temporal behavioral activities of employees. The proposed approach identifies insiders with an Accuracy of $\mathbf{9 3. 7 7 \%}$, Precision of $\mathbf{9 3. 8 9 \%}$, Recall of $\mathbf{9 9. 6 7 \%}$, F1-score of $\mathbf{9 6. 6 9 \%}$, which shows the effectiveness of the quantum inspired approach to tackle this threat.