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

arXiv:2206.10955 · IEEE Transactions on Information Forensics and Security

Adversarial Reconfigurable Intelligent Surface Against Physical Layer Key Generation

Zhuangkun Wei, Bin Li, Weisi Guo

Published Jan 1, 2023• 33 citations
arXivHTMLPDF

Abstract

The development of reconfigurable intelligent surfaces (RIS) has recently advanced the research of physical layer security (PLS). Beneficial impacts of RIS include but are not limited to offering a new degree-of-freedom (DoF) for key-less PLS optimization, and increasing channel randomness for physical layer secret key generation (PL-SKG). However, there is a lack of research studying how adversarial RIS can be used to attack and obtain legitimate secret keys generated by PL-SKG. In this work, we show an Eve-controlled adversarial RIS (Eve-RIS), by inserting into the legitimate channel a random and reciprocal channel, can partially reconstruct the secret keys from the legitimate PL-SKG process. To operationalize this concept, we design Eve-RIS schemes against two PL-SKG techniques used: (i) the CSI-based PL-SKG, and (ii) the two-way cross multiplication based PL-SKG. The channel probing at Eve-RIS is realized by compressed sensing designs with a small number of radio-frequency (RF) chains. Then, the optimal RIS phase is obtained by maximizing the Eve-RIS inserted deceiving channel. Our analysis and results show that even with a passive RIS, our proposed Eve-RIS can achieve a high key match rate with legitimate users, and is resistant to most of the current defensive approaches. This means the novel Eve-RIS provides a new eavesdropping threat on PL-SKG, which can spur new research areas to counter adversarial RIS attacks.

DOI
10.1109/tifs.2023.3266705

Related papers

Is Adversarial RIS Harmful to Physical Layer Secret Key Generation

Yi Fang, Huici Wu, Zhiqing Wei +1 more

Reconfigurable intelligent surface (RIS) is acknowledged as a promising technique in improving physical layer (PHY) secret key generation (SKG) performance by enhancing the randomness of radio electromagnetic environment. However, RIS can also be adopted by illegal eavesdropper to destroy the channel reciprocity between key-sharing parties. Thus, there is a tradeoff between the enhanced randomness and the decreased channel reciprocity with the application of RIS. This paper aims to answer the question that whether the adversarial RIS is harmful to SKG by studying the SKG performance in the RIS scenario where Alice and Eve each applies a RIS. To this end, the closed-form expression for the upper bound of secret key rate (SKR) and expressions for the autocorrelation coefficients of two probing channel samples within a coherent duration are provided. Results show that the relative positions of Eve and adversarial RIS to Alice and Bob play a key role in determining whether the adversarial RIS is harmful or beneficial to the PHY SKG system.

Globecom•Dec 4, 2023•2 citations

Explainable Adversarial Learning Framework on Physical Layer Secret Keys Combating Malicious Reconfigurable Intelligent Surface

Zhuangkun Wei, Wenxiu Hu, Zhang, Junqing +2 more

Reconfigurable intelligent surfaces (RIS) can both help and hinder the physical layer secret key generation (PL-SKG) of communications systems. Whilst a legitimate RIS can yield beneficial impacts, including increased channel randomness to enhance PL-SKG, a malicious RIS can poison legitimate channels and crack almost all existing PL-SKGs. In this work, we propose an adversarial learning framework that addresses Man-in-the-middle RIS (MITM-RIS) eavesdropping which can exist between legitimate parties, namely Alice and Bob. First, the theoretical mutual information gap between legitimate pairs and MITM-RIS is deduced. From this, Alice and Bob leverage adversarial learning to learn a common feature space that assures no mutual information overlap with MITM-RIS. Next, to explain the trained legitimate common feature generator, we aid signal processing interpretation of black-box neural networks using a symbolic explainable AI (xAI) representation. These symbolic terms of dominant neurons aid the engineering of feature designs and the validation of the learned common feature space. Simulation results show that our proposed adversarial learning- and symbolic-based PL-SKGs can achieve high key agreement rates between legitimate users, and is further resistant to an MITM-RIS Eve with the full knowledge of legitimate feature generation (NNs or formulas). This therefore paves the way to secure wireless communications with untrusted reflective devices in future 6G.

arXiv:2402.06663•Feb 6, 2024•0 citations

Secret Key Rate Upper-bound for Reconfigurable Intelligent Surface-combined System under Spoofing

Zhuangkun Wei, Liang Wang, Weisi Guo

Reconfigurable intelligent surfaces (RIS) have been shown to improve the secret key rate (SKR) for physical layer secret key generation (PL-SKG), by using the programmable phase shifts to increase reciprocal channel entropy. Most current studies consider the role of RIS on passive eavesdroppers (Eves) and overlook active attackers, especially the pilot spoofing attacks (PSA). For PSA in PL-SKG setups, this is implemented by Eve sending an amplified pilot sequence simultaneously with legitimate user Alice. With the increase of the spoofing amplifying factor, the channel probing results at Bob and Eve become similar, thereby enabling Eve to generate shared secret key with Bob. In this work, we analyze how RIS can positively or negatively affect the PL-SKG under pilot spoofing. To do so, we theoretically express the legitimate and spoofing SKRs in terms of the RIS phase shifts. Leveraging this, the closed-form theoretical upper bounds of both legitimate and spoofing SKRs are deduced, which lead to two further findings. First, the legitimate SKR upper-bound does not vary with RIS phase shift vector, but reduces drastically with the increase of the spoofing amplifying factor. This suggests the limited effect of RIS against PL-SKG spoofing, since the legitimate SKR has a hard limit, which cannot be surpassed by adjusting RIS phase and reflecting power, but can even be 0 with properly assigned spoofing amplifying factor. Second, the spoofing SKR upper-bound shows a large gap from the non-optimized SKR, which indicates a potential for RIS phase optimization.

IEEE Vehicular Technology Conference•Sep 1, 2022•7 citations

Explainable Adversarial Learning Framework on Physical Layer Key Generation Combating Malicious Reconfigurable Intelligent Surface

Zhuangkun Wei, Wenxiu Hu, Junqing Zhang +2 more

Reconfigurable intelligent surfaces (RIS) can both help and hinder the physical layer secret key generation (PL-SKG) of communications systems. Whilst a legitimate RIS can yield beneficial impacts, including increased channel randomness to enhance PL-SKG, a malicious RIS can poison legitimate channels and crack almost all existing PL-SKGs. In this work, we propose an adversarial learning framework that addresses Man-in-the-middle RIS (MITM-RIS) eavesdropping which can exist between legitimate parties, namely Alice and Bob. First, the theoretical mutual information gap between legitimate pairs and MITM-RIS is deduced. From this, Alice and Bob leverage adversarial learning to learn a common feature space that assures no mutual information overlap with MITM-RIS. Next, to explain the trained legitimate common feature generator, we aid signal processing interpretation of black-box neural networks using a symbolic explainable AI (xAI) representation. These symbolic terms of dominant neurons aid the engineering of feature designs and the validation of the learned common feature space. Simulation results show that our proposed adversarial learning- and symbolic-based PL-SKGs can achieve high key agreement rates between legitimate users, and is further resistant to an MITM-RIS Eve with the full knowledge of legitimate feature generation (NNs or formulas). This therefore paves the way to secure wireless communications with untrusted reflective devices in future 6G.

IEEE Transactions on Wireless Communications•Jan 28, 2025•9 citations

Reconfigurable Intelligent Surface-Assisted PHY Key Generation: A Tradeoff Analysis

Huici Wu, Yi Fang, Zhiqing Wei +2 more

Reconfigurable intelligent surface (RIS) has been widely studied as a promising approach in improving secret key rate (SKR) for physical layer (PHY) secret key generation (SKG). Existing RIS-assisted PHY SKG studies mainly focus on improving the SKR by optimizing RIS phase. However, RIS can also be exploited by active attackers to contaminate the channel probing. As a result, there is a trade-off between the increased randomness and the improved correlation between legitimate channels and eavesdropping channels in a RIS-assisted SKG system under channel contamination. In this paper, we study the impact of RIS on PHY SKG by analyzing the SKR performance. A closed-form expression for the lower-bound of secret key capacity is obtained. Theoretical analysis shows that gains of legitimate channel, eavesdropping channel, and RIS-induced channels are key parameters in determining whether the assisting RIS is beneficial or harmful to the PHY SKG. Further, a SKR maximization problem under channel contamination is proposed and solved. Results show that the assisting RIS is definitely harmful to the SKG when the gain ratio of legitimate channel and eavesdropping channel is reduced with the leverage of RIS, while it can significantly improve the SKR with an optimized RIS phase when the gain ratio is improved with the help of RIS.

IEEE Transactions on Vehicular Technology•Dec 26, 2024•2 citations

A Multi-Eavesdropper Scheme Against RIS Secured LoS-Dominated Channel

Zhuangkun Wei, Weisi Guo, Bin Li

Reconfigurable intelligent surface (RIS) has been shown as a promising technique to increase the channel randomness for secret key generation (SKG) in low-entropy channels (e.g., static or line-of-sight (LoS)), without small-scale fading. In this letter, we show that even with the aid of RIS, collaborative eavesdroppers (Eves) can still estimate the legitimate Alice-Bob channel and erode their secret key rates (SKRs), since the RIS induced randomness is also reflected in the Eves’ observations. Conditioned on Eves’ observations, if the entropy of RIS-combined legitimate channel is zero, Eves are able to estimate it and its secret key. Leveraging this, we design a multi-Eve scheme against the RIS-secured LoS dominated scenarios, by using the multiple Eves’ observations to reconstruct the RIS-combined legitimate channel. We further deduce a closed-form secret key leakage rate under our designed multi-Eve scheme, and demonstrate the results via simulations.

IEEE Communications Letters•Apr 11, 2022•15 citations
The development of reconfigurable intelligent surfaces (RIS) has recently advanced the research of physical layer security (PLS). Beneficial impacts of RIS include but are not limited to offering a new degree-of-freedom (DoF) for key-less PLS optimization, and increasing channel randomness for physical layer secret key generation (PL-SKG). However, there is a lack of research studying how adversarial RIS can be used to attack and obtain legitimate secret keys generated by PL-SKG. In this work, we show an Eve-controlled adversarial RIS (Eve-RIS), by inserting into the legitimate channel a random and reciprocal channel, can partially reconstruct the secret keys from the legitimate PL-SKG process. To operationalize this concept, we design Eve-RIS schemes against two PL-SKG techniques used: (i) the CSI-based PL-SKG, and (ii) the two-way cross multiplication based PL-SKG. The channel probing at Eve-RIS is realized by compressed sensing designs with a small number of radio-frequency (RF) chains. Then, the optimal RIS phase is obtained by maximizing the Eve-RIS inserted deceiving channel. Our analysis and results show that even with a passive RIS, our proposed Eve-RIS can achieve a high key match rate with legitimate users, and is resistant to most of the current defensive approaches. This means the novel Eve-RIS provides a new eavesdropping threat on PL-SKG, which can spur new research areas to counter adversarial RIS attacks.
Reconfigurable intelligent surface (RIS) is acknowledged as a promising technique in improving physical layer (PHY) secret key generation (SKG) performance by enhancing the randomness of radio electromagnetic environment. However, RIS can also be adopted by illegal eavesdropper to destroy the channel reciprocity between key-sharing parties. Thus, there is a tradeoff between the enhanced randomness and the decreased channel reciprocity with the application of RIS. This paper aims to answer the question that whether the adversarial RIS is harmful to SKG by studying the SKG performance in the RIS scenario where Alice and Eve each applies a RIS. To this end, the closed-form expression for the upper bound of secret key rate (SKR) and expressions for the autocorrelation coefficients of two probing channel samples within a coherent duration are provided. Results show that the relative positions of Eve and adversarial RIS to Alice and Bob play a key role in determining whether the adversarial RIS is harmful or beneficial to the PHY SKG system.
Reconfigurable intelligent surfaces (RIS) can both help and hinder the physical layer secret key generation (PL-SKG) of communications systems. Whilst a legitimate RIS can yield beneficial impacts, including increased channel randomness to enhance PL-SKG, a malicious RIS can poison legitimate channels and crack almost all existing PL-SKGs. In this work, we propose an adversarial learning framework that addresses Man-in-the-middle RIS (MITM-RIS) eavesdropping which can exist between legitimate parties, namely Alice and Bob. First, the theoretical mutual information gap between legitimate pairs and MITM-RIS is deduced. From this, Alice and Bob leverage adversarial learning to learn a common feature space that assures no mutual information overlap with MITM-RIS. Next, to explain the trained legitimate common feature generator, we aid signal processing interpretation of black-box neural networks using a symbolic explainable AI (xAI) representation. These symbolic terms of dominant neurons aid the engineering of feature designs and the validation of the learned common feature space. Simulation results show that our proposed adversarial learning- and symbolic-based PL-SKGs can achieve high key agreement rates between legitimate users, and is further resistant to an MITM-RIS Eve with the full knowledge of legitimate feature generation (NNs or formulas). This therefore paves the way to secure wireless communications with untrusted reflective devices in future 6G.
Reconfigurable intelligent surfaces (RIS) have been shown to improve the secret key rate (SKR) for physical layer secret key generation (PL-SKG), by using the programmable phase shifts to increase reciprocal channel entropy. Most current studies consider the role of RIS on passive eavesdroppers (Eves) and overlook active attackers, especially the pilot spoofing attacks (PSA). For PSA in PL-SKG setups, this is implemented by Eve sending an amplified pilot sequence simultaneously with legitimate user Alice. With the increase of the spoofing amplifying factor, the channel probing results at Bob and Eve become similar, thereby enabling Eve to generate shared secret key with Bob. In this work, we analyze how RIS can positively or negatively affect the PL-SKG under pilot spoofing. To do so, we theoretically express the legitimate and spoofing SKRs in terms of the RIS phase shifts. Leveraging this, the closed-form theoretical upper bounds of both legitimate and spoofing SKRs are deduced, which lead to two further findings. First, the legitimate SKR upper-bound does not vary with RIS phase shift vector, but reduces drastically with the increase of the spoofing amplifying factor. This suggests the limited effect of RIS against PL-SKG spoofing, since the legitimate SKR has a hard limit, which cannot be surpassed by adjusting RIS phase and reflecting power, but can even be 0 with properly assigned spoofing amplifying factor. Second, the spoofing SKR upper-bound shows a large gap from the non-optimized SKR, which indicates a potential for RIS phase optimization.
Reconfigurable intelligent surfaces (RIS) can both help and hinder the physical layer secret key generation (PL-SKG) of communications systems. Whilst a legitimate RIS can yield beneficial impacts, including increased channel randomness to enhance PL-SKG, a malicious RIS can poison legitimate channels and crack almost all existing PL-SKGs. In this work, we propose an adversarial learning framework that addresses Man-in-the-middle RIS (MITM-RIS) eavesdropping which can exist between legitimate parties, namely Alice and Bob. First, the theoretical mutual information gap between legitimate pairs and MITM-RIS is deduced. From this, Alice and Bob leverage adversarial learning to learn a common feature space that assures no mutual information overlap with MITM-RIS. Next, to explain the trained legitimate common feature generator, we aid signal processing interpretation of black-box neural networks using a symbolic explainable AI (xAI) representation. These symbolic terms of dominant neurons aid the engineering of feature designs and the validation of the learned common feature space. Simulation results show that our proposed adversarial learning- and symbolic-based PL-SKGs can achieve high key agreement rates between legitimate users, and is further resistant to an MITM-RIS Eve with the full knowledge of legitimate feature generation (NNs or formulas). This therefore paves the way to secure wireless communications with untrusted reflective devices in future 6G.
Reconfigurable intelligent surface (RIS) has been widely studied as a promising approach in improving secret key rate (SKR) for physical layer (PHY) secret key generation (SKG). Existing RIS-assisted PHY SKG studies mainly focus on improving the SKR by optimizing RIS phase. However, RIS can also be exploited by active attackers to contaminate the channel probing. As a result, there is a trade-off between the increased randomness and the improved correlation between legitimate channels and eavesdropping channels in a RIS-assisted SKG system under channel contamination. In this paper, we study the impact of RIS on PHY SKG by analyzing the SKR performance. A closed-form expression for the lower-bound of secret key capacity is obtained. Theoretical analysis shows that gains of legitimate channel, eavesdropping channel, and RIS-induced channels are key parameters in determining whether the assisting RIS is beneficial or harmful to the PHY SKG. Further, a SKR maximization problem under channel contamination is proposed and solved. Results show that the assisting RIS is definitely harmful to the SKG when the gain ratio of legitimate channel and eavesdropping channel is reduced with the leverage of RIS, while it can significantly improve the SKR with an optimized RIS phase when the gain ratio is improved with the help of RIS.
Reconfigurable intelligent surface (RIS) has been shown as a promising technique to increase the channel randomness for secret key generation (SKG) in low-entropy channels (e.g., static or line-of-sight (LoS)), without small-scale fading. In this letter, we show that even with the aid of RIS, collaborative eavesdroppers (Eves) can still estimate the legitimate Alice-Bob channel and erode their secret key rates (SKRs), since the RIS induced randomness is also reflected in the Eves’ observations. Conditioned on Eves’ observations, if the entropy of RIS-combined legitimate channel is zero, Eves are able to estimate it and its secret key. Leveraging this, we design a multi-Eve scheme against the RIS-secured LoS dominated scenarios, by using the multiple Eves’ observations to reconstruct the RIS-combined legitimate channel. We further deduce a closed-form secret key leakage rate under our designed multi-Eve scheme, and demonstrate the results via simulations.