Channel State Information (CSI) Fingerprinting

The broader physical-layer security discipline that identifies wireless transmitters by analyzing hardware-specific imperfections in emitted waveforms. CSI fingerprinting is a specialized subset that leverages fine-grained channel measurements rather than transient turn-on signatures.

• Uses I/Q imbalance, oscillator drift, and power amplifier non-linearity as discriminators
• Operates at the waveform level, independent of cryptographic identifiers like MAC addresses
• Vulnerable to spoofing attacks using high-quality software-defined radios

A hardware security primitive that exploits inherent manufacturing variations in silicon to produce a unique, unclonable device fingerprint. PUFs and CSI fingerprinting share the foundational principle that physical randomness can serve as an unforgeable identity.

• Generates challenge-response pairs derived from sub-threshold voltage differences
• Used for cryptographic key generation without storing secrets in non-volatile memory
• Complements CSI-based authentication in multi-factor device attestation schemes

A metric that measures the cosine of the angle between two non-zero vectors in an inner product space. In CSI fingerprinting, it is frequently used to compare feature embeddings extracted from channel estimates to determine if two measurements originate from the same location or device.

• Range: -1 (opposite) to 1 (identical), with 0 indicating orthogonality
• Invariant to vector magnitude, focusing purely on directional similarity
• Computationally efficient for real-time matching against large fingerprint databases

A search technique that finds points in a vector space that are approximately closest to a query point, trading marginal accuracy for orders-of-magnitude speed improvements. Essential for scaling CSI fingerprinting systems to billion-scale device databases.

• Employs techniques like locality-sensitive hashing (LSH) and graph-based indices
• Reduces query latency from linear scan O(n) to sub-linear or logarithmic time
• Libraries like FAISS provide GPU-accelerated implementations for production deployment

A neural architecture containing two or more identical subnetworks with shared weights, trained to learn a similarity metric between input pairs. Widely adopted in CSI fingerprinting for one-shot device authentication where training data per device is scarce.

• Outputs a distance score rather than a class label, enabling open-set recognition
• Trained with contrastive loss or triplet loss to minimize intra-class distance and maximize inter-class distance
• Enables authentication of previously unseen devices without retraining the entire model

The systematic evaluation of machine learning models against inputs deliberately crafted to cause misclassification. In CSI fingerprinting, attackers may use channel manipulation or signal replay with subtle distortions to impersonate legitimate devices.

• Tests resilience against evasion attacks where adversarial perturbations are added to channel estimates
• Evaluates spatial spoofing scenarios where an attacker mimics a trusted location's CSI profile
• Critical for certifying CSI-based authentication systems for high-assurance environments