Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

IEEE Vehicular Technology Conference

GAN-Assisted Secret Key Generation Against Eavesdropping In Dynamic Indoor LiFi Networks

Elmahedi Mahalal, Muhammad Ismail, Zi-Yang Wu, Mostafa M. Fouda, Zubair Md. Fadlullah

Published Oct 7, 2024• 1 citation
Publisher

Abstract

This paper explores the vulnerability of wireless secret key generation (WSKG) to eavesdropping in a dynamic indoor light-fidelity (LiFi) network. It analyzes the channel impulse response (CIR) similarities of two moving user equipments (UEs) across scenarios with two, four, and eight UEs. We observe that as the number of UEs increases, the similarity in CIR also rises, due to the proximal movement patterns among UEs. Specifically, the similarity rate peaks at 70% when eight UEs enter the room; it then drops to 24% during the wandering phase and rises again to 80% as UEs exit the room. Consequently, an eavesdropper among the eight UEs is able to generate 27% of a legitimate UE’s secret key, it significantly reduces the key’s complexity, decreasing the number of possible keys that need to be tested to break the encryption and making it easier to predict the remainder of the key. To mitigate this issue, we introduce a novel approach that utilizes a generative adversarial network (GAN) to artificially manipulate the CIR, thereby reducing the effectiveness of eavesdropping by adding noise into the observed CIR. This method effectively reduces the CIR similarity to a negligible 1%, thus ensuring the integrity of WSKG against eavesdropping threats.

DOI
10.1109/vtc2024-fall63153.2024.10757826

Related papers

GAN-Based Artificial Noise Generation Against Eavesdropping for Wireless Secret Key Generation in Dynamic Indoor LiFi Networks

Elmahedi Mahalal, Eslam Hasan, Muhammad Ismail +4 more

This paper investigates the vulnerability of wireless secret key generation (WSKG) to eavesdropping by analyzing the channel impulse response (CIR) between a legitimate user and an eavesdropper under various scenarios in dynamic indoor light fidelity (LiFi) networks. These scenarios include (a) different user densities (2, 4, and 8 users), (b) fields of view (FoVs) of$30^\circ$,$60^\circ$, and$90^\circ$, and (c) various room layouts. Results show that higher user densities increase downlink CIR similarity, as users' movement traces become closer. For instance, with eight users and a$30^\circ$FoV, CIR similarity peaks at over 85%–95% during entering and exiting stages and remains no less than 40% during wandering. Consequently, an eavesdropper can generate a key with 27%–37% similarity to the legitimate user's key. To mitigate this threat, we propose a novel defense using a generative adversarial network (GAN) trained with crafted uplink CIRs. GANs model the complex statistical properties of legitimate CIRs and generate synthetic noise that mimics environmental and system characteristics without replicating real user CIRs. This prevents eavesdroppers from extracting useful information and avoids privacy concerns linked to handling actual CIR data. Furthermore, traditional reversed CIR methods are less effective in dynamic environments, where conditions change rapidly and are easier to reverse-engineer. Our GAN-generated noise, applied within the defense zone of the legitimate user, reduces CIR similarities from up to 95% to approximately 1%, effectively nullifying key leakage. These findings highlight the potential of GAN-based noise to significantly enhance WSKG security in dynamic indoor LiFi networks.

IEEE Open Journal of Vehicular Technology•Jan 1, 2026•1 citation

Characterization of Secret Key Generation in 5G+ Indoor Mobile LiFi Networks

Elmahedi Mahalal, Muhammad Ismail, Zi-Yang Wu +1 more

This paper investigates wireless secret key generation in 5G+ indoor LiFi networks operating in the visible light band in the downlink and the infrared band in the uplink. Unlike the existing research, this paper aims to characterize the impact of indoor user mobility on the ability to generate identical secret keys between the user and the base station. Hence, we first generate mobility traces that reflect realistic human behavior in an indoor setup to capture the sensitivity of the optical channels to random blockage and transceiver-misorientation events associated with user mobility. Next, we generate the corresponding channel impulse response and adopt guard band quantization and information reconciliation to extract identical uplink and downlink keys. Then, we examine a set of spatial and temporal features related to the channel probing rate, key disagreement rate, and key generation rate, along with their statistical distributions. Further, we study the randomness of the key following the NIST randomness tests. Our study demonstrated that the handover strategy significantly impacts the randomness of the key and its generation rate.

Globecom•Dec 4, 2022•6 citations

Concept Drift Aware Wireless Key Generation in Dynamic LiFi Networks

Elmahedi Mahalal, Eslam Hasan, Muhammad Ismail +4 more

This paper studies the generation of cryptographic keys from wireless channels in light-fidelity (LiFi) networks. Unlike existing studies, we account for several practical considerations (a) realistic indoor multi-user mobility scenarios, (b) non-ideal channel reciprocity given the unique characteristics of the downlink visible light (VL) and uplink infrared (IR) channels, (c) different room occupancy levels, (d) different room layouts, and (e) different receivers’ field-of-view (FoV). Since general channel models in dynamic LiFi networks are inaccurate, we propose a novel deep learning-based framework to generate secret keys with minimal key disagreement rate (KDR) and maximal key generation rate (KGR). However, we find that wireless channels in LiFi networks exhibit different statistical behaviors under various conditions, leading to concept drift in the deep learning model. As a result, key generation suffers from (a) a deterioration in KDR and KGR up to 29% and 38%, respectively, and (b) failing the NIST randomness test. To enable a concept drift aware framework, we propose an adaptive learning strategy using the similarity of channel probability density functions and the mix-of-experts ensemble method. Results show our adaptive learning strategy can achieve stable performance that passes the NIST randomness test and achieves 8% KDR and 89 bits/s KGR for a case of study with 60° FoV.

IEEE Open Journal of the Communications Society•Dec 31, 2024•14 citations

Robust Deep Learning-Based Secret Key Generation in Dynamic LiFi Networks Against Concept Drift

Elmahedi Mahalal, Muhammad Ismail, Zi-Yang Wu +3 more

This paper explores secret key generation in 5G and beyond LiFi networks using visible light in the downlink and infrared in the uplink. Unlike the existing works, we focus on a realistic indoor environment with multi-user mobility. Given inaccuracies in high-frequency channel models, we introduce the first deep learning model that combines the channel probing and quantization phases to generate initial secret keys with a minimal key disagreement rate (KDR) of 16% between the uplink and downlink, leading to a key generation rate (KGR) of 79 bits/s after information reconciliation. We show that LiFi channel statistics suffer from concept drifts with user density changes in the room. This increases the KDR by 28% - 44% and the generated keys fail to pass the NIST randomness tests. As a countermeasure, we introduce a voting ensemble model that mitigates concept drifts, maintaining a stable 16% KDR, 79 bits/s KGR, and passing NIST tests, despite the varying user densities.

IEEE Consumer Communications and Networking Conference•Jan 6, 2024•4 citations

Key Generation Technique Based on Channel Characteristics for MIMO-OFDM Wireless Communication Systems

Dinh Van Linh, Van Yem Vu

Dynamic secret key generation from wireless channel characteristics is a promising technique for physical layer security. One of the important issues in this field is extending the secret key’s length while preserving its uniformity and randomness. This paper proposes a key generation method based on time-varying and the reciprocity of wireless channels for Multiple-Input Multiple-Output Orthogonal Frequency Division Multiplexing (MIMO-OFDM) wireless communication systems. In the proposed technique, the complex impulse response (CIR) of the estimated channel state information (CSI) is extracted, and a quantization algorithm is designed to convert the maximum peaks of the CIR into secret key bits. The effectiveness of the proposed key generation technique is assessed in terms of the randomness of the produced key bits with different key lengths by using a statistical test suite of the National Institute of Standards and Technology (NIST). The proposed technique is employed in the MIMO-OFDM systems with different modulation schemes through Additive White Gaussian Noise (AWGN) and Rayleigh channels. The simulation results show that the secret keys with various key lengths generated from the proposed technique for the MIMO-OFDM systems guarantee randomness. Moreover, the proposed CSI-based key generation technique provides better effectiveness in terms of security when compared to some existing techniques.

IEEE Access•Jan 1, 2023•27 citations

Deep Learning for Hardware-Impaired Wireless Secret Key Generation with Man-in-the-Middle Attacks

Mehdi Letafati, Hamid Behroozi, Babak Hossein Khalaj +1 more

Wireless secret key generation (WSKG) allows efficient key agreement protocols for securing the sixth generation (6G) wireless networks. Nevertheless, due to external adversaries or internal impairments, WSKG schemes might become vulner-able during the randomness distillation, where the legitimate nodes try to observe their source of common randomness. In this paper, we investigate the WSKG scheme with legitimate parties suffering from hardware impairments (HIs), while an active adversary acts as a man-in-the-middle (MiM) via injecting fake pilot signals. We first utilize randomized pilots to overcome the MiM. We also leverage the concept of recurrent neural networks (RNNs) to further enhance the randomness distillation. More specifically, the long short-term memory networks (LSTMs)-as a well-established type of RNNs-are implemented to learn the long-term dependencies between the observations of legitimate parties. The achievable secret key rate (SKR) and the impact of MiM on system's performance are analyzed. Our numerical results verify the performance gain of our proposed learning-based approach compared with the state-of-the-art methods and provide useful insights on system design. We show that our RNN-based approach achieves 30% and 15% improvement in terms of observation mismatches compared with the naïve scheme and the fully-connected benchmarks, respectively.

Globecom•Dec 1, 2021•29 citations
This paper explores the vulnerability of wireless secret key generation (WSKG) to eavesdropping in a dynamic indoor light-fidelity (LiFi) network. It analyzes the channel impulse response (CIR) similarities of two moving user equipments (UEs) across scenarios with two, four, and eight UEs. We observe that as the number of UEs increases, the similarity in CIR also rises, due to the proximal movement patterns among UEs. Specifically, the similarity rate peaks at 70% when eight UEs enter the room; it then drops to 24% during the wandering phase and rises again to 80% as UEs exit the room. Consequently, an eavesdropper among the eight UEs is able to generate 27% of a legitimate UE’s secret key, it significantly reduces the key’s complexity, decreasing the number of possible keys that need to be tested to break the encryption and making it easier to predict the remainder of the key. To mitigate this issue, we introduce a novel approach that utilizes a generative adversarial network (GAN) to artificially manipulate the CIR, thereby reducing the effectiveness of eavesdropping by adding noise into the observed CIR. This method effectively reduces the CIR similarity to a negligible 1%, thus ensuring the integrity of WSKG against eavesdropping threats.
This paper investigates the vulnerability of wireless secret key generation (WSKG) to eavesdropping by analyzing the channel impulse response (CIR) between a legitimate user and an eavesdropper under various scenarios in dynamic indoor light fidelity (LiFi) networks. These scenarios include (a) different user densities (2, 4, and 8 users), (b) fields of view (FoVs) of30∘30^\circ30∘,60∘60^\circ60∘, and90∘90^\circ90∘, and (c) various room layouts. Results show that higher user densities increase downlink CIR similarity, as users' movement traces become closer. For instance, with eight users and a30∘30^\circ30∘FoV, CIR similarity peaks at over 85%–95% during entering and exiting stages and remains no less than 40% during wandering. Consequently, an eavesdropper can generate a key with 27%–37% similarity to the legitimate user's key. To mitigate this threat, we propose a novel defense using a generative adversarial network (GAN) trained with crafted uplink CIRs. GANs model the complex statistical properties of legitimate CIRs and generate synthetic noise that mimics environmental and system characteristics without replicating real user CIRs. This prevents eavesdroppers from extracting useful information and avoids privacy concerns linked to handling actual CIR data. Furthermore, traditional reversed CIR methods are less effective in dynamic environments, where conditions change rapidly and are easier to reverse-engineer. Our GAN-generated noise, applied within the defense zone of the legitimate user, reduces CIR similarities from up to 95% to approximately 1%, effectively nullifying key leakage. These findings highlight the potential of GAN-based noise to significantly enhance WSKG security in dynamic indoor LiFi networks.
This paper investigates wireless secret key generation in 5G+ indoor LiFi networks operating in the visible light band in the downlink and the infrared band in the uplink. Unlike the existing research, this paper aims to characterize the impact of indoor user mobility on the ability to generate identical secret keys between the user and the base station. Hence, we first generate mobility traces that reflect realistic human behavior in an indoor setup to capture the sensitivity of the optical channels to random blockage and transceiver-misorientation events associated with user mobility. Next, we generate the corresponding channel impulse response and adopt guard band quantization and information reconciliation to extract identical uplink and downlink keys. Then, we examine a set of spatial and temporal features related to the channel probing rate, key disagreement rate, and key generation rate, along with their statistical distributions. Further, we study the randomness of the key following the NIST randomness tests. Our study demonstrated that the handover strategy significantly impacts the randomness of the key and its generation rate.
This paper studies the generation of cryptographic keys from wireless channels in light-fidelity (LiFi) networks. Unlike existing studies, we account for several practical considerations (a) realistic indoor multi-user mobility scenarios, (b) non-ideal channel reciprocity given the unique characteristics of the downlink visible light (VL) and uplink infrared (IR) channels, (c) different room occupancy levels, (d) different room layouts, and (e) different receivers’ field-of-view (FoV). Since general channel models in dynamic LiFi networks are inaccurate, we propose a novel deep learning-based framework to generate secret keys with minimal key disagreement rate (KDR) and maximal key generation rate (KGR). However, we find that wireless channels in LiFi networks exhibit different statistical behaviors under various conditions, leading to concept drift in the deep learning model. As a result, key generation suffers from (a) a deterioration in KDR and KGR up to 29% and 38%, respectively, and (b) failing the NIST randomness test. To enable a concept drift aware framework, we propose an adaptive learning strategy using the similarity of channel probability density functions and the mix-of-experts ensemble method. Results show our adaptive learning strategy can achieve stable performance that passes the NIST randomness test and achieves 8% KDR and 89 bits/s KGR for a case of study with 60° FoV.
This paper explores secret key generation in 5G and beyond LiFi networks using visible light in the downlink and infrared in the uplink. Unlike the existing works, we focus on a realistic indoor environment with multi-user mobility. Given inaccuracies in high-frequency channel models, we introduce the first deep learning model that combines the channel probing and quantization phases to generate initial secret keys with a minimal key disagreement rate (KDR) of 16% between the uplink and downlink, leading to a key generation rate (KGR) of 79 bits/s after information reconciliation. We show that LiFi channel statistics suffer from concept drifts with user density changes in the room. This increases the KDR by 28% - 44% and the generated keys fail to pass the NIST randomness tests. As a countermeasure, we introduce a voting ensemble model that mitigates concept drifts, maintaining a stable 16% KDR, 79 bits/s KGR, and passing NIST tests, despite the varying user densities.
Dynamic secret key generation from wireless channel characteristics is a promising technique for physical layer security. One of the important issues in this field is extending the secret key’s length while preserving its uniformity and randomness. This paper proposes a key generation method based on time-varying and the reciprocity of wireless channels for Multiple-Input Multiple-Output Orthogonal Frequency Division Multiplexing (MIMO-OFDM) wireless communication systems. In the proposed technique, the complex impulse response (CIR) of the estimated channel state information (CSI) is extracted, and a quantization algorithm is designed to convert the maximum peaks of the CIR into secret key bits. The effectiveness of the proposed key generation technique is assessed in terms of the randomness of the produced key bits with different key lengths by using a statistical test suite of the National Institute of Standards and Technology (NIST). The proposed technique is employed in the MIMO-OFDM systems with different modulation schemes through Additive White Gaussian Noise (AWGN) and Rayleigh channels. The simulation results show that the secret keys with various key lengths generated from the proposed technique for the MIMO-OFDM systems guarantee randomness. Moreover, the proposed CSI-based key generation technique provides better effectiveness in terms of security when compared to some existing techniques.
Wireless secret key generation (WSKG) allows efficient key agreement protocols for securing the sixth generation (6G) wireless networks. Nevertheless, due to external adversaries or internal impairments, WSKG schemes might become vulner-able during the randomness distillation, where the legitimate nodes try to observe their source of common randomness. In this paper, we investigate the WSKG scheme with legitimate parties suffering from hardware impairments (HIs), while an active adversary acts as a man-in-the-middle (MiM) via injecting fake pilot signals. We first utilize randomized pilots to overcome the MiM. We also leverage the concept of recurrent neural networks (RNNs) to further enhance the randomness distillation. More specifically, the long short-term memory networks (LSTMs)-as a well-established type of RNNs-are implemented to learn the long-term dependencies between the observations of legitimate parties. The achievable secret key rate (SKR) and the impact of MiM on system's performance are analyzed. Our numerical results verify the performance gain of our proposed learning-based approach compared with the state-of-the-art methods and provide useful insights on system design. We show that our RNN-based approach achieves 30% and 15% improvement in terms of observation mismatches compared with the naïve scheme and the fully-connected benchmarks, respectively.