How Feasible are Passive Network Attacks on 5G Networks and Beyond? A Survey

The gist The survey examines the feasibility of passive network attacks in 5G and beyond (B5G/6G) networks, with emphasis on two major categories: information extraction (system identification, website and application fingerprinting) and geolocation (user identification and position tracking).

Introduction to Passive Network Attacks

Passive Network Attacks are defined as Attacks that monitor or intercept network traffic without altering it, aiming to gather information covertly <ref:2512.20622#pg8>. These attacks rely on Communication side-channels which exploit indirect information to infer sensitive data without direct access to content <ref:2512.20622#pg4>. The main objectives of packet trace analysis include traffic classification (see Fig. 2 for an illustration), user identification and localization, website fingerprinting, application recognition, and data-content inference <ref:2512.20622#pg6>. These attacks are well documented and reproducible in existing wireless and mobile systems, including short-range networks (IEEE 802.11) and, to a lesser extent, LTE <ref:2512.20622#pg4>. Current evidence suggests that while such attacks remain theoretically possible in 5G, their practical execution is significantly constrained by directional beamforming, high-frequency propagation characteristics, and encryption mechanisms <ref:2512.20622#pg4>.

Key Types of Passive Network Attacks

The survey focuses on two major categories of passive network attacks: information extraction and geolocation <ref:2512.20622#pg4>. Information extraction includes system identification, website fingerprinting, and application fingerprinting <ref:2512.20622#pg4>, while geolocation involves user identification and position tracking <ref:2512.20622#pg4>. Specific attack targets include Website Fingerprinting (WF), Video Fingerprinting (VF), Smart Home and Behavioral Inference, De-anonymization in Privacy-Enhancing Networks, User Identification, Tracking (TR), and Position Tracking (CPM) <ref:2512.20622#pg3>.

Feasibility Analysis in 5G and Beyond

The feasibility of PNAs is classified into four levels based on reproducibility under 5G/B5G constraints <ref:2512.20622#pg14>. Class 1, Not Reproducible / Not Applicable, includes attacks that rely on data that is fundamentally inaccessible under the 5G security model. Class 4, Easily Reproducible, is noted as not being met by any known attack in 5G, as attacks require SDR hardware, custom protocol decoders, or non-commodity data collection pipelines. Class 3, Reproducible with efforts, involves attacks that are feasible on 5G testbeds or small scale live networks where techniques such as decoding control plane messages or extracting side channel features have been shown to work.

Constraints and Challenges for Attack Reproducibility

Several factors constrain the practical execution of these attacks in 5G environments. These constraints include physical layer limitations, protocol layer barriers, and resource barriers. Physical Layer Constraints involve The Beamforming Barrier, where an eavesdropper must overcome the spatial filtering inherent in the transmitter’s antenna array and faces an SNR Wall for Demodulation where leakage SNR often falls below the Shannon limit for reliable decoding. Protocol Layer Barriers include The RNTI Scrambling Hurdle, where a passive sniffer must brute-force the RNTI to decode a single DCI message, leading to a search space explosion. Resource Barriers involve High-Bandwidth Acquisition Costs, as capturing 5G spectrum requires SDRs costing upwards of ten thousand dollars, and the Lack of 'Promiscuous' Software Stacks for 5G compared to LTE makes sniffing more difficult.

Conclusion on Future Feasibility

The classification reveals that "most classical attacks such as packet-sequence traffic classification and flow-based fingerprinting are no longer reproducible in 5G, since the required visibility is blocked by MAC encryption, multiplexing, and beamforming. The paper concludes that 5G does not eliminate PNAs entirely, but it confines them to research-grade or well-equipped adversaries. Looking forward, the move toward B5G/6G is expected to further reduce feasibility due to the absence of public sniffers, increasing hardware costs, and added obfuscation mechanisms".

The survey systematically analyzed the current landscape of passive network attacks in the context of 5G and beyond. The comparative analysis shows that most classical attacks such as packet-sequence traffic classification and flow-based fingerprinting are no longer reproducible in 5G, since the required visibility is blocked by MAC encryption, multiplexing, and beamforming. Additional obstacles include the need for specialized SDR hardware, the limited maturity of opensource 5G sniffing tools, and the directional nature of high-frequency links. Nevertheless, recent studies demonstrate that a subset of PNAs exploiting control-plane leaks (e.g., through DCI or RRC analysis) and physical-layer side channels (e.g., RF fingerprinting) remain feasible in 5G, though only with non-trivial expertise and equipment.

How it works

The survey investigates the feasibility of PNAs by classifying existing works into four levels of feasibility based on input types: packet sequences, flows, or physical/control-plane data. Class 1 attacks are Not Reproducible / Not Applicable, as they rely on data that is fundamentally inaccessible under the 5G security model. Class 4 attacks are deemed Easily Reproducible, but no known attack meets these criteria in 5G, as even demonstrations require SDR hardware or custom decoders.

Conclusion on Future Feasibility

The survey concludes that "most classical attacks such as packet-sequence traffic classification and flow-based fingerprinting are no longer reproducible in 5G, since the required visibility is blocked by MAC encryption, multiplexing, and beamforming. The paper suggests that 5G does not eliminate PNAs entirely, but it confines them to research-grade or well-equipped adversaries. Looking forward, the move toward B5G/6G is expected to further reduce feasibility due to the absence of public sniffers, increasing hardware costs, and added obfuscation mechanisms".

1. The AI system can be improved to incorporate control-plane and physical-layer leakages for more robust identification, as a third distinct family is increasingly present; relying on control-plane or physical-layer features (see Fig. 6 for a statistic). This allows the system to exploit control-plane messages or physical-layer features such as DCI scheduling, uplink feedback, or RF characteristics which are the most practical input sources for modern PNAs.

2. The system can be enhanced to handle noisy data and device variability by using models that are robust to these conditions, as prior work shows that classification accuracy is rather reduced when a ML model trained on a given device or a given brand is used on a different one. This suggests developing models like probabilistic modeling tools such as Hidden Markov Models and finite state machines for application classification under link-layer encryption, where traditional flow-based ML fails due to lack of packet headers.

3. The AI system can be designed with spatial awareness to account for directional communication constraints, as 5G NR (especially in FR2/mmWave) utilizes massive MIMO and beamforming to direct energy toward the legitimate user (UE). This means the system's inference capabilities must incorporate knowledge of spatial filtering inherent in the transmitter’s antenna array, potentially leading to a more accurate assessment of signal reception feasibility based on distance and beam alignment.

4. The system can be adapted for deployment in 5G/B5G environments by using specialized decoding techniques, as no complete, public end-to-end 5G passive monitoring stack is yet available. This implies the AI needs to be trained or designed to process data derived from custom-developed tools or patched forks of LTE-era decoders that extract telemetry and scheduling metadata from PDCCH/DCI messages.

Privacy concerns around 5G, the latest generation of mobile networks, are growing, with fears that its deployment may increase exposure to privacy risks. This perception is largely driven by the use of denser deployments of small antenna systems, which enable highly accurate data collection at higher speeds and closer proximity to mobile users. At the same time, 5G's unique radio communication features can make the reproduction of known network attacks more challenging. In particular, passive network attacks, which do not involve direct interaction with the target network and are therefore nearly impossible to detect, remain a pressing concern. Such attacks can reveal sensitive information about users, their devices, and active applications, which may then be exploited through known vulnerabilities or spear-phishing schemes. This survey examines the feasibility of passive network attacks in 5G and beyond (B5G/6G) networks, with emphasis on two major categories: information extraction (system identification, website and application fingerprinting) and geolocation (user identification and position tracking). These attacks are well documented and reproducible in existing wireless and mobile systems, including short-range networks (IEEE 802.11) and, to a lesser extent, LTE. Current evidence suggests that while such attacks remain theoretically possible in 5G, their practical execution is significantly constrained by directional beamforming, high-frequency propagation characteristics, and encryption mechanisms. For B5G and early 6G networks, the lack of public tools and high hardware cost currently renders these attacks infeasible in practice, which highlights a critical gap in our understanding of future network threat models.