AI-Driven Threat Detection and Response Using Quantum Cryptography

AI-Driven Threat Detection and Response Using Quantum Cryptography

Prof. Mohini A. Thorat

Department of Computer Engineering, JSPMs Jayawantrao Sawant College of Engineering, Pune

Anushka Kumbhar

Department of Computer Engineering, JSPMs Jayawantrao Sawant College of Engineering, Pune

Janhvi Chavan

Department of Computer Engineering, JSPMs Jayawantrao Sawant College of Engineering, Pune

Mithilesh Kadam

Department of Computer Engineering, JSPMs Jayawantrao Sawant College of Engineering, Pune

Aditya Shinde

Department of Computer Engineering, JSPMs Jayawantrao Sawant College of Engineering, Pune

Abstract – The rapid advancement of quantum computing poses a signicant threat to conventional cryptographic systems that rely on computational complexity for security [3]. Quan-tum Key Distribution (QKD) offers a secure communication mechanism based on the principles of quantum mechanics, enabling the detection of unauthorized interception attempts [1]. However, practical QKD deployments remain vulnerable to channel noise, hardware imperfections, and sophisticated eavesdropping attacks [5], [6]. This paper presents an AI-driven threat detection and response framework that combines BB84 Quantum Key Distribution, IBM Qiskit simulation, and machine learning-based anomaly detection [9]. A high-delity quantum simulator is developed using IBM Qiskit and AerSimulator to model realistic quantum communication channels with depo-larizing noise. A Random Forest classier trained on 50,000 simulated transmission sequences identies malicious activities by analyzing Quantum Bit Error Rate (QBER), noise levels, sifted key lengths, and engineered heuristic features [7]. Experimental analysis demonstrates high detection accuracy while maintaining real- time operational capabilities. The proposed system further incorporates automated mitigation techniques including Privacy Amplication and E91 protocol migration to enhance quantum network resilience [2]. The results demonstrate that integrating articial intelligence with quantum cryptography signicantly improves threat detection, response efciency, and overall com- munication security.

Index TermsQuantum Cryptography, QKD, BB84, Random Forest, IBM Qiskit, Threat Detection, Anomaly Detection, Secu-rity Operations Center

1. Introduction

   The increasing computational power of modern systems and the emergence of quantum computing technologies have raised serious concerns regarding the long-term security of traditional cryptographic algorithms [10]. Widely deployed encryption schemes such as RSA and Elliptic Curve Cryptography rely on mathematical problems that are computationally infeasible

   for classical computers. However, quantum algorithms such as Shors Algorithm have demonstrated the theoretical capa-bility to solve these problems exponentially faster, potentially rendering current cryptographic infrastructures obsolete [3].

   Quantum Key Distribution (QKD) has emerged as a promis-ing solution to this challenge. Unlike classical encryption systems that depend on computational hardness, QKD de-rives its security from the fundamental principles of quantum mechanics [4]. The BB84 protocol, introduced by Bennett and Brassard, enables two communicating parties to establish a shared secret key while ensuring that any eavesdropping attempt introduces detectable disturbances into the quantum channel [1].

   Although QKD offers theoretically unconditional security, practical implementations face several challenges. Environ- mental noise, transmission losses, hardware imperfections, and malicious attacks can signicantly affect communication reliability [6]. Among these threats, intercept-resend attacks remain one of the most important security concerns. In such attacks, an adversary intercepts transmitted photons, performs measurements, and forwards replacement photons to the re- ceiver. Due to the No-Cloning Theorem and wave-function collapse, these actions introduce measurable errors into the communication channel [5].

   Machine learning techniques have been proposed to en-hance anomaly detection in network security systems

   [8],[11]. Integrating machine learning with quantum cryptography creates an intelligent layer capable of distinguishing malicious interception from natural channel noise, enabling automated and adaptive threat response [12].

2. Literature Survey

   BB84 established the foundation of quantum cryptography by enabling secure key exchange with eavesdropping detection [1]. E91 later improved security through entanglement-based communication [2]. Research by Scarani et al. highlighted the importance of Quantum Bit Error Rate (QBER) in evaluating channel security [5], while Pirandola et al. discussed practical deployment challenges in quantum networks [6].

   Machine learning techniques have been widely applied in cybersecurity [8]. Random Forest, introduced by Breiman, offers high accuracy and robustness against overtting, making it suitable for anomaly detection [7]. Although substantial research exists in both quantum cryptography and machine learning, few studies integrate quantum simulation, threat de- tection, visualization, and automated response within a single framework [11], [12].

   A. Research Gap

   • Limited intelligent threat detection in QKD systems [5].

   • Dependence on static QBER thresholds [6].

   • Lack of automated mitigation mechanisms.

   • Minimal integration of AI, quantum simulation, and SOC monitoring [8].

3. Proposed System

   The proposed system presents an intelligent Security Op- erations Center (SOC) architecture designed for real-time monitoring, detection, and mitigation of threats in Quantum Key Distribution (QKD) networks. The framework integrates quantum cryptographic simulation, machine learning-based anomaly detection, automated response mechanisms, and in- teractive visualization to provide end-to-end security manage- ment.

   1. System Objectives

      The primary objectives of the proposed system are:

      • To simulate realistic BB84 Quantum Key Distribution channels using IBM Qiskit [9].

      • To model environmental noise and intercept-resend at- tacks using quantum mechanical principles [10].

      • To develop an intelligent threat detection framework using machine learning [7].

      • To distinguish malicious activities from normal channel degradation [8].

      • To provide real-time visualization and monitoring capa- bilities.

      • To automate threat response and channel protection mechanisms.

   2. System Architecture

   The proposed architecture consists of the following layers:

   1. Quantum Simulation Layer

   2. Data Generation and Feature Extraction Layer

   3. Machine Learning Inference Layer

   4. Security Operations Center Dashboard

   5. Automated Mitigation Layer

   The Quantum Simulation Layer generates quantum com- munication data using BB84 protocol operations [1]. The generated data is processed to compute Quantum Bit Error Rate (QBER), noise levels, and sifted key lengths [5]. These values are thn forwarded to the Machine Learning Layer for threat classication [7]. Finally, the dashboard visualizes system status and initiates mitigation procedures when threats are detected.

   Feature Extraction (QBER, Noise, Key Length)

   Random Forest Threat Detection

   SOC Dashboard (Real-Time Monitoring)

Automated Mitigation (Privacy Amplication / E91)

BB84 Quantum Simulator (IBM Qiskit)

Fig. 1. Architecture of the Proposed AI-Driven QKD Threat Detection Framework.

The system consists of ve sequential layers: Quantum Simulation, Feature Extraction, Machine Learning Inference, SOC Monitoring, and Automated Mitigation.

The framework consists of ve layers as shown in Fig. 1: Quantum Simulation, Feature Extraction, Machine Learning Inference, SOC Monitoring, and Automated Mitigation. Quan- tum communication data generated using BB84 [1] is pro- cessed into security metrics and classied by a Random Forest model [7]. Threats detected by the classier trigger automated countermeasures.

1. Methodology

   1. Phase 1: Quantum Simulation

      BB84 communication is simulated using IBM Qiskit and AerSimulator [9]. Alice and Bob exchange quantum states through noisy channels, generating realistic communication records [1], [10].

   2. Phase 2: Attack Injection

      Intercept-resend attacks are introduced by allowing an ad- versary to measure and retransmit photons. Incorrect basis selection increases QBER and produces identiable attack patterns [4], [5].

   3. Phase 3: Feature Extraction

      The extracted features include:

      • QBER (Quantum Bit Error Rate) [5]

      • Noise Level (Depolarizing noise parameter) [9]

      • Sifted Key Length

      • Effective Key Rate

      • Eve Contribution (engineered heuristic feature)

        EveContribution = max(0, QBER 0.66 × NoiseLevel) (1)

   4. Phase 4: Threat Detection

      A Random Forest classier is trained using 50,000 simu- lated transmission records [7]. The model classies commu- nication sessions as secure or compromised. Random Forest was selected for its robustness against overtting, high accu- racy on high-dimensional data, and interpretability of feature importance [7], [8].

   5. Phase 5: Automated Mitigation

   Threats trigger Privacy Amplication and migration to the E91 protocol to maintain communication security [2], [6].

   100

   98

   Score (%)

   96

   94

   92

   90

   Random Forest Classier Performance

   95.8

   95.3

   94.9

   93.5

   Accuracy

   Precision

   Recall

   F1-Score

   Metric

2. Experimental Results and Discussion

   The experimental environment consisted of the following components:

   • IBM Qiskit for BB84 quantum simulation [9]

   • Qiskit AerSimulator for noisy quantum execution [9]

   • Python and FastAPI backend services

   • Scikit-learn Random Forest Classier [7]

   • React-based Security Operations Center Dashboard

   • Dataset consisting of 50,000 quantum transmission records

   Each simulation generated communication metrics includ- ing Quantum Bit Error Rate (QBER), noise level, sifted key length, effective key rate, and engineered heuristic features.

   1. Performance Metrics

      To evaluate the effectiveness of the proposed model, stan- dard machine learning performance metrics were employed: Accuracy, Precision, Recall, and F1-Score [7], [8].

      Accuracy measures overall prediction correctness, while Precision evaluates the percentage of correctly identied at- tacks among all detected attacks. Recall measures the models ability to identify actual attacks, and F1-Score provides a balanced evaluation of Precision and Recall.

   2. Classication Results

      The Random Forest classier demonstrated strong perfor- mance in distinguishing malicious interception attempts from normal environmental noise.

      Metric	Value
      Accuracy	93.5%
      Precision	95.8%
      Recall	94.9%
      F1-Score	95.3%

      TABLE I

      Classification Results of Random Forest Classifier

      As shown in Table 1, the Random Forest classier achieved high accuracy, precision, recall, and F1-score, demonstrating its effectiveness in distinguishing malicious attacks from nor- mal environmental noise [7].

      Fig. 2. Comparative performance of the Random Forest classier

      Fig. 2 illustrates the comparative performance of the Ran- dom Forest classier across the four evaluation metrics, high- lighting consistently high values for accuracy, precision, recall, and F1-score.

   3. Impact of QBER on Threat Detection

      Quantum Bit Error Rate (QBER) plays a critical role in determining channel security [5]. Under normal operating conditions, QBER remained within acceptable limits caused by environmental disturbances and transmission imperfections. However, when intercept-resend attacks were introduced, sig- nicant increases in QBER were observed [4].

      Normal Channel Intercept-Resend Attack QBER Threshold (11%)

      QBER (%)

      40

      20

      0

      0 5 10 15 20

      Simulation Trial

      Fig. 3. QBER values under normal channel conditions versus intercept-resend attack conditions across 20 simulation trials.

      The machine learning model successfully learned these behavioral patterns and utilized them for reliable attack clas- sication, as illustrated in Fig. 3 [5], [7].

   4. Effectiveness of Feature Engineering

      A custom heuristic feature was introduced to isolate attacker-induced disturbances from natural channel noise:

      EveContribution = max(0, QBER 0.66 × NoiseLevel) (2)

      The engineered feature signicantly improved classication performance by removing expected environmental depolariza- tion effects and highlighting anomalies caused by malicious interference [5]. Experimental observations indicated that this feature contributed substantially to reducing false alarm rates while improving attack detection accuracy [7].

   5. Automated Mitigation Analysis

   Upon detection of a high-probability threat, the Security Operations Center automatically initiated mitigation proce- dures [6]. The rst mitigation stage employed Privacy Am- plication to reduce attacker knowledge regarding generated keys. Although this process reduced effective key generation rates, it restored communication security. Migration to the E91 protocol was subsequently triggered for sessions under sustained attack [2].

3. Advantages of the Proposed System

   • Real-Time Monitoring: Provides continuous monitoring of quantum communication channels [8].

   • Accurate Attack Detection: Accurately detects intercept-resend attacks using machine learning [7].

   • Noise Differentiation: Differentiates malicious activities from normal environmental noise using engineered fea- tures [5].

   • Realistic Simulation: Utilizes realistic quantum simula- tions through IBM Qiskit [9].

   • Automated Response: Triggers countermeasures auto- matically without manual intervention [6].

   • Scalability: The modular architecture allows extension to additional QKD protocols and hardware platforms [4].

4. Future Scope

   Future enhancements can further improve the capabilities of the proposed framework:

   • Integration with real quantum hardware available through IBM Quantum platforms [9].

   • Support for additional QKD protocols beyond BB84 and E91 [2], [4].

   • Application of deep learning models for advanced threat intelligence [11], [12].

   • Deployment in enterprise quantum network testbeds for real-world validation [6].

   • Federated learning approaches to enable collaborative threat detection across distributed QKD nodes [12].

5. Conclusion

This paper presented an AI-driven threat detection and response framework for securing Quantum Key Distribution networks against malicious interception attempts [1], [3]. The proposed architecture integrates IBM Qiskit-based BB84 simulation [9], Random Forest machine learning classication [7], real-time Security Operations Center monitoring, and automated mitigation mechanisms [2].

A dataset consisting of 50,000 simulated transmission se- quences was utilized to train and evaluate the detection model.

Experimental results demonstrated strong performance with an accuracy of 93.5%, precision of 95.8%, recall of 94.9%, and F1-score of 95.3% [7].

The integration of Privacy Amplication and E91 protocol migration further enhanced communication security by en- abling automated defensive responses [2], [6]. The overall re- sults indicate that combining articial intelligence with quan- tum cryptography provides an effective and scalable solution for protecting future quantum communication infrastructures [5], [8].

Acknowledgment

The authors express their sincere gratitude to Prof. Mohini

A. Thorat, Department of Computer Engineering, JSPMs Jayawantrao Sawant College of Engineering, Pune, for her valuable guidance, encouragement, and continuous support throughout the development of this project and research work.

References

1. C. H. Bennett and G. Brassard, Quantum Cryptography: Public Key Distribution and Coin Tossing, Proceedings of IEEE International Conference on Computers, Systems and Signal Processing, Bangalore, India, 1984.

2. A. K. Ekert, Quantum Cryptography Based on Bells Theorem,

   Physical Review Letters, vol. 67, no. 6, pp. 661663, 1991.

3. P. W. Shor, Algorithms for Quantum Computation: Discrete Logarithms and Factoring, Proceedings of the 35th Annual Symposium on Founda- tions of Computer Science, pp. 124134, 1994.

4. N. Gisin, G. Ribordy, W. Tittel and H. Zbinden, Quantum Cryptogra- phy, Reviews of Modern Physics, vol. 74, no. 1, pp. 145195, 2002.

5. V. Scarani et al., The Security of Practical Quantum Key Distribution,

   Reviews of Modern Physics, vol. 81, no. 3, pp. 13011350, 2009.

6. S. Pirandola et al., Advances in Quantum Cryptography, Advances in Optics and Photonics, vol. 12, no. 4, pp. 10121236, 2020.

7. L. Breiman, Random Forests, Machine Learning, vol. 45, no. 1, pp. 532, 2001.

8. R. Sommer and V. Paxson, Outside the Closed World: On Using Machine Learning for Network Intrusion Detection, IEEE Symposium on Security and Privacy, 2010.

9. H. Abraham et al., Qiskit: An Open-Source Framework for Quantum Computing, IBM Research, 2024.

10. M. Nielsen and I. Chuang, Quantum Computation and Quantum Infor- mation, Cambridge University Press, 2010.

11. H. J. Liao, C. H. R. Lin, Y. C. Lin, and K. Y. Tung, Intrusion Detection System: A Comprehensive Review, Journal of Network and Computer Applications, vol. 36, no. 1, pp. 1624, 2013.

12. S. Garcia, M. Grill, J. Stiborek, and A. Zunino, An Empirical Compar- ison of Botnet Detection Methods, Computers & Security, vol. 45, pp. 100123, 2014.

13. H.-K. Lo, M. Curty, and K. Tamaki, Secure Quantum Key Distribution,

    Nature Photonics, vol. 8, pp. 595604, 2014.

14. M. Lucamarini, Z. L. Yuan, J. F. Dynes, and A. J. Shields, Overcoming the Rate-Distance Limit of Quantum Key Distribution Without Quantum Repeaters, Nature, vol. 557, pp. 400403, 2018.

15. F. Xu, X. Ma, Q. Zhang, H.-K. Lo, and J.-W. Pan, Secure Quantum Key Distribution with Realistic Devices, Reviews of Modern Physics, vol. 92, no. 2, p. 025002, 2020.

16. R. Alle´aume et al., Using Quantum Key Distribution for Cryptographic Purposes: A Survey, Theoretical Computer Science, vol. 560, pp. 62 81, 2014.

17. I. Vagniluca et al., Efcient Time-Bin Encoding for Practical High- Dimensional Quantum Key Distribution, Physical Review Applied, vol. 14, p. 014051, 2020.