VeriPHY: Unified Verification & Authentication

• VeriPHY is a suite of methodologies that combines formal verification, runtime monitoring, and deep learning-based authentication to secure interactions at the physical layer.
• It applies rigorous techniques from differential dynamic logic to lattice theory, delivering provable correctness in robotics, network control, and V2X communication.
• By integrating formal methods with practical runtime safeguards and advanced signal processing, VeriPHY ensures robust security while managing performance-compliance trade-offs.

VeriPHY is a technical term for a suite of research methodologies and implementations that center on formal verification, runtime enforcement, and, more recently, deep learning–based authentication in physical layer networks, with notable deployments spanning from robotics safety monitors to physical layer security in next-generation wireless environments. Across its research lineage, VeriPHY targets assurance at the level where signals, protocols, and hardware interact—using a combination of formal methods, practical runtime monitoring, and physical-layer signal processing/authentication—to deliver provable correctness and robust security guarantees. The following sections survey the principal variants, methodologies, and applications associated with VeriPHY, drawing on key papers and results from the field.

1. Formal Methods and End-to-End Verification

VeriPHY's foundational methodology is rooted in differential dynamic logic (dL) and hybrid program modeling, first implemented for verifying safety and liveness in ground robots (Bohrer et al., 2019). Systems are specified by a hybrid model alternating between a nondeterministic controller (choosing next waypoint, curvature, speed bounds, acceleration command) and plant dynamics, which advance physical state according to controlled differential equations:

$$t := 0; \ \{ \ \frac{dx}{dt} = v(y-1), \ \frac{dy}{dt} = -vx, \ \frac{dv}{dt} = a, \ \frac{dt}{dt} = 1 \} \qquad \text{[with domain constraint:} \ t \leq T, v \geq 0 \text{]}$$

Controllers are subject to feasibility (Feas) and admissibility (Go) constraints—such as “annular” tolerance for waypoints and achievable speed/acceleration boundaries—collectively ensuring the system's assumptions hold in a real-world deployment.

2. Runtime Monitors and Machine Code Enforcement

A critical innovation in the VeriPHY approach is using formal proofs to synthesize runtime monitors which are automatically compiled into machine code (Bohrer et al., 2019). Two monitor classes are generated:

• Controller Monitor: Checks each controller output (waypoint, speed, curvature, acceleration) to confirm compliance with proven bounds.
• Plant Monitor: Validates that sensor readings and physical state evolution remain within the admissible invariants of the verified model.

Fallback actions (e.g., verified braking logic) are invoked if either monitor detects violation. The toolchain ensures that the formal guarantees carry over down to the compiled binary, so that, in the presence of untrusted controllers or environments, end-to-end safety/liveness is still enforced at runtime.

3. Expressive Network Verification via Lattice Theory

For extending real-time network verification, VeriPHY benefits from lattice-theoretical methodologies for packet equivalence class (PEC) partitioning, as described in (Horn et al., 2019). The #PEC algorithm constructs a meet–semilattice ordering over match conditions, organizing packet headers into minimal, unique atomic predicates. This formalism enables:

• Precise representation of complex forwarding rules (ip-tables, ranges, multi-field negations)
• Elimination of empty PECs, avoiding false positives/negatives in error detection (e.g., shadowed rules)
• Efficient verification with up to 10× speedup compared to previous methods (e.g., Veriflow, atomic predicate–based approaches)
• Scalable performance and minimal storage overhead in large network environments

This framework is compatible with various “element types” (e.g., ip_prefix, tbv<N>, range, set<T>) and integrates seamlessly with VeriPHY’s architecture for correctness and performance.

4. Physical Layer Abstraction and V2X Modelling

VeriPHY is also applied to abstracting the physical layer in direct V2X communication scenarios (Zhuofei et al., 2022). Here, a single parameter α (“implementation loss”) captures the difference between measured throughput and ideal Shannon channel capacity:

$$\Psi_e(\theta) \approx \alpha \cdot \Psi_s(\gamma_\text{th}(\theta)), \qquad \Psi_s(\gamma_\text{th}) = B \cdot \log_2(1 + \gamma_\text{th})$$

Calibrated through least-squares fitting on reference measurements, α allows the methodology to generalize physical layer metrics (PER vs SINR curves) across various configurations and environments, dramatically reducing simulation complexity while preserving network-level evaluation accuracy. Cross-layer dependencies (MAC procedures, transmission times) are folded into this abstraction, yielding precise throughput and performance estimates for both IEEE 802.11p and LTE-V2X technologies.

5. Deep Learning-Based Physical Layer Authentication

Recent advances extend VeriPHY to secure wireless communication at the physical layer by embedding device-specific signatures into I/Q transmissions using steganography and deep learning (Robinson et al., 11 Aug 2025). Core methodologies include:

• Signature Generation: Each user equipment (UE) generates pseudo-random signatures by sampling $n_{el}$ values from a Gaussian Mixture Model (GMM), with uniqueness enforced by a minimum Kolmogorov–Smirnov distance ε between users' GMMs.
• Embedding & Detection: Signatures are embedded over I/Q samples in real time (e.g., intervals of 1–20 ms) and detected at the 5G gNB using convolutional neural networks such as SENet and VGG16, which analyze 2D signal representations.
• Stealth Mode: A scaling or masking operation ensures that signature alterations are nearly indistinguishable from clean 5G signals, confirmed via energy distribution analysis, while maintaining detection accuracy above 93%.
• Performance: Reported detection rates range from 93–100% with inference times as low as 6.5 ms (for 1 ms signature intervals), demonstrating feasibility for high-throughput, low-latency environments.

This combination of physical layer authentication (PLA) and deep neural networks provides secure, cryptography-free early-layer blocking of unauthorized access, particularly relevant for emerging 5G, IoT, and V2X systems.

6. Trade-offs, Applications, and Reliability

VeriPHY’s implementations and deployments reveal inherent trade-offs between performance and compliance. In robotics applications, more aggressive controllers may trigger monitor fallback more often, reducing operational agility but maintaining safety (Bohrer et al., 2019). In physical signal authentication, stealth signature generation may slightly reduce detection accuracy but prevent attack vectors that exploit observable anomalies (Robinson et al., 11 Aug 2025). In V2X abstraction, the use of α allows generalization at the cost of some granularity, yet network-level accuracy is preserved (Zhuofei et al., 2022).

Practical use cases include:

• Safety-critical robotics and autonomous ground vehicles
• Real-time network verification in software-defined networking and data centers
• Secure wireless infrastructure, including base station–level blocking in 5G and encrypted physical layer transmission

7. Future Directions and Broader Implications

Across its variants, VeriPHY is poised for further development in areas such as mutual PLA (bilateral device and base station authentication), cross-layer integrated security (combining formal verification and physical layer signal methods), and broader class of cyber-physical systems. The foundational work in lattice-theoretical partitioning and end-to-end formal proof synthesis offers deep connections to formal methods, control theory, and applied machine learning, suggesting continued relevance in both academic research and real-world deployments. Future research anticipates extension to satellite, IoT, and mission-critical domains, as well as further optimization in detection architectures and robust signature embedding under evolving channel and attack models.

---

Overall, VeriPHY encapsulates a set of distinct, rigorously formalized, and experimentally validated techniques in physical layer verification and authentication, uniting the principles of runtime safety monitoring, expressive network verification, cross-layer abstraction, and deep learning–based signal authentication.