Battlefield 5G: Dual-PKI and TPM-Based UE Attestation for Tactical 5G Standalone Networks

The paper presents Battlefield 5G, a pre-authentication framework for tactical 5G Standalone (SA) networks that addresses a gap in standard 5G-AKA: "The standardized 5G Authentication and Key Agreement (5G-AKA) authenticates a subscriber credential stored on a Universal Subscriber Identity Model (USIM) but does not authenticate the physical device that holds that credential or verify its boot state. This gap is critical in tactical deployments where user equipment may be captured, modified, returned to service, or used with transplanted subscriber credentials."

The system combines dual X.509 device-certificate checks with Trusted Platform Module (TPM) -based boot attestation before standard registration is accepted. Specifically, the design places "an outer certificate challenge on the 5G base-station called gNB, an independent inner certificate challenge on the Access and Mobility Management Function (AMF) in the 5G core network, and a TPM PCR (Platform Configuration Register) quote verified by an attestation proxy on the 5G core network side."

The implementation uses a gNodeB (gNB) side Radio Resource Control (RRC) forwarding gate and an AMF-side save-and-replay mechanism that "enable multi-round certificate and attestation challenge-response exchanges to be inserted into the registration path without modifying any 3GPP Non-Access Stratum (NAS) message structures or adding new NAS message types. The prototype was built by extending the Radio Access Network of the Software Radio System (srsRAN), gNB, User Equipment of the Software Radio System (srsUE) and Open5GS in a B210-based Universal Radio Peripheral (USRP) testbed with a hardware TPM 2.0 in the UE."

The dual-certificate design is "inspired by the National Security Agency (NSA) Commercial Solutions for Classified (CSfC) key-management guidance, which uses separate Outer and Inner Certificate Authorities (CA) so that the same certificate cannot authenticate between both protection layers. The system uses two independent certificate hierarchies: Outer Root CA → Outer CA → UE Outer Cert trusted only by the gNB, and Inner Root CA → Inner CA → UE Inner Cert trusted only by the AMF. Thus, the same credential cannot satisfy both checkpoints."

The protocol inserts three pre-authentication checks before standard 5G-AKA processing: "Check 1 is an outer device-certificate challenge issued by the gNB after RRC setup and before NAS forwarding; Check 2 is an independent inner device-certificate challenge issued by the AMF before the saved Registration Request is processed; Check 3 is a challenge to the TPM attestation issued by the AMF and verified by an attestation proxy." Only after passing all three checks does the AMF replay the saved Registration Request and continue standard 5G registration.

The threat model considers an adversary that "can capture a legitimate UE, extract or reuse its USIM, modify firmware or boot components, deploy an SDR-based rogue platform, observe over-the-air messages, and replay previously captured certificate or attestation responses. The system enforces four goals: (1) device identity, where the UE must prove possession of an enrolled device private key; (2) checkpoint separation, where compromise of one certificate checkpoint does not bypass all admission checks; (3) boot integrity, where TPM PCRs must match an enrolled baseline; and (4) replay resistance, where captured responses cannot satisfy fresh challenges."

The evaluation used a three-node 5G SA testbed with USRP B210 radios, Open5GS core, and a TPM 2.0 device in the UE. The prototype blocks SIM-transplant, rogue-certificate, firmware-tampering, and replay attacks. Specifically: SIM-transplant and rogue-certificate attacks are blocked at the gNB because the UE lacks the required outer device certificate/private key or the certificate is self-signed or not issued by the trusted Outer CA. Firmware tampering is blocked after TPM verification because the quoted PCR values do not match the enrolled baseline. Replay attacks are blocked by nonce freshness at the replayed checkpoint; certificate responses and TPM quotes cannot be reused across sessions.

Latency results show that "Across six trials, Battlefield 5G increases average onboarding latency from 1886 ms to 2260 ms, adding 373.4 ms of pre-authentication overhead while preserving standard 5G-AKA, security mode, and packet data unit (PDU) session procedures. The latency breakdown shows: The Outer CA exchange averages 74.9 ms, but the local OpenSSL verification at the gNB takes only 1.0 ms, showing that certificate processing is not the bottleneck. The Inner CA exchange averages 49.2 ms... The RTT TPM attestation dominates the added delay at 235.7 ms, or approximately 79% of the 298.5 ms AMF-side pre-authentication latency. The system also adds 3609 bytes of signaling per connection: 1100 bytes for the Outer CA exchange, 1100 bytes for the Inner CA exchange, and 1409 bytes for the TPM attestation exchange."

The security analysis concludes that The proposed pipeline enforces three complementary properties before network admission: device identity in the gNB, independent device identity in the AMF, and boot integrity in the attestation proxy. The paper notes limitations: "The current prototype verifies PCRs 0–7, which cover early boot integrity but not runtime user-space changes. Runtime integrity can be added using PCR 10 and IMA event-log verification. In addition, the current enrollment database is a flat file." These are identified as ongoing work.

The conclusion states: "These results show that certificate-based device identity and hardware-rooted boot integrity verification can be made a part of the 5G access decision for tactical networks. In contested environments where compromised or unauthorized devices must not be admitted to the network, this pre-admission layer provides a practical and deployable mechanism on commodity 5G hardware without changing 3GPP NAS message structures or replacing standard 5G-AKA, security mode, and PDU session procedures."

Improvements to AI Systems:

1. AI-Driven Threat-Adaptive Admission Controller: Build an AI layer that dynamically adjusts the pre-authentication checks based on real-time threat intelligence. The AI can learn from attack patterns (e.g., SIM-transplant attempts, rogue certificates) to tighten or relax specific checkpoints (e.g., require additional TPM PCR quotes or extend nonce lifetimes) without human intervention, reducing latency during low-risk periods while hardening during active attacks.

2. Predictive Latency Optimizer for Pre-Authentication: Use a machine learning model to predict the optimal timing and parallelization of the three checks (Outer CA, Inner CA, TPM attestation). The AI can reorder or overlap non-dependent operations (e.g., pre-fetching TPM quotes while certificate verification runs) to reduce the 373 ms overhead, targeting a 20–30% latency reduction while maintaining security guarantees.

3. Anomaly Detection for TPM PCR Baseline Drift: Implement an AI-based anomaly detector that monitors TPM PCR values across multiple enrollment sessions. Instead of static baselines, the AI learns the normal variance of PCR 0–7 and flags subtle firmware tampering that produces near-baseline but non-identical quotes, catching attacks that static matching might miss.

4. Intelligent Replay-Attack Classifier: Train a classifier on historical replay attempts (e.g., reused nonces, repeated certificate responses) to distinguish benign retransmissions from malicious replays. The AI can automatically update the nonce freshness policy and checkpoint challenge generation to preemptively invalidate known replay vectors before they reach the gNB or AMF.

5. Self-Healing Enrollment Database Manager: Replace the flat-file enrollment database with an AI-managed, self-healing store that detects and corrects inconsistencies (e.g., duplicated certificates, stale PCR baselines). The AI can automatically re-enroll devices with updated baselines after legitimate firmware updates, reducing false rejections while maintaining security.

6. Cross-Layer Certificate Trust Analyzer: Use a graph neural network to analyze the dual-certificate hierarchies (Outer and Inner CAs) for subtle trust relationships or misconfigurations. The AI can identify if a compromised intermediate CA could inadvertently satisfy both checkpoints, and proactively suggest certificate revocation or re-issuance before an attack exploits it.

7. Adaptive Attestation Proxy Load Balancer: Deploy an AI that predicts TPM attestation request bursts (e.g., during mass device onboarding) and dynamically scales attestation proxy resources or prioritizes critical devices. This prevents the 235 ms TPM bottleneck from becoming a denial-of-service vector in tactical scenarios.

What the Improved AI System Can Do:

• Autonomously harden or relax 5G admission in real time based on threat level, reducing overhead by up to 30% in safe zones while maintaining full security in contested areas.

• Detect and block novel firmware tampering that deviates subtly from enrolled PCR baselines, which static checks would miss.

• Prevent replay attacks proactively by learning and invalidating attack patterns before they are replayed at checkpoints.

• Self-maintain the enrollment database without human intervention, ensuring devices with legitimate updates are not falsely rejected.

• Optimize resource allocation for attestation proxies to prevent latency spikes during mass onboarding or coordinated attacks.

• Provide a security posture dashboard that explains why specific devices were admitted or blocked, aiding tactical operators in decision-making.

The standardized 5G Authentication and Key Agreement (5G-AKA) authenticates a subscriber credential stored on a Universal Subscriber Identity Model (USIM) but does not authenticate the physical device that holds that credential or verify its boot state. This gap is significant in tactical 5G deployments, where user equipment may be captured, modified, returned to service, or used with transplanted subscriber credentials. We present Battlefield 5G, a pre-authentication framework for 5G Standalone networks that combines dual X.509 device-certificate checks with Trusted Platform Module (TPM) -based boot attestation before standard registration is accepted. The design places an outer certificate challenge on the 5G base-station called gNB, an independent inner certificate challenge on the Access and Mobility Management Function (AMF) in the 5G core network, and a TPM PCR (Platform Configuration Register) quote verified by an attestation proxy on the 5G core network side. A gNodeB (gNB) side Radio Resource Control (RRC) forwarding gate and an AMF-side save-and-replay mechanism enable multi-round certificate and attestation challenge-response exchanges to be inserted into the registration path without modifying any 3GPP Non-Access Stratum (NAS) message structures or adding new NAS message types. We implement these capabilities by extending the Radio Access Network of the Software Radio System (srsRAN), gNB, User Equipment of the Software Radio System (srsUE) and Open5GS in a B210-based Universal Radio Peripheral (USRP) testbed with a hardware TPM 2.0 in the UE. The prototype blocks SIM-transplant, rogue-certificate, firmware-tampering, and replay attacks. Across six trials, Battlefield 5G increases average onboarding latency from 1886 ms to 2260 ms, adding 373.4 ms of pre-authentication overhead while preserving standard 5G-AKA, security mode, and packet data unit (PDU) session procedures.