Wireless Sensing Security Through Deceptive-Feature Training

The patent applies preliminary anti-action by training machine learning models with deceptive features before deployment. The models are pre-trained to recognize and reject signals containing deceptive features, which attackers might use for passive sensing. This proactive measure prevents unauthorized sensing by establishing defense mechanisms in advance, rather than reacting to attacks after they occur.

The patent converts the harmful effect of deceptive features into a beneficial security mechanism. By intentionally introducing deceptive features during training, the system learns to distinguish between legitimate and illegitimate signals. The deceptive features, which could potentially be exploited by attackers, are instead used to strengthen the model's ability to detect and reject unauthorized passive sensing attempts.

The patent applies preliminary action by preparing training datasets with deceptive features in advance. Instead of dealing with complex security challenges during deployment, the system pre-trains models with various deceptive scenarios. This preliminary preparation simplifies the deployment phase, as the models are already equipped to handle security threats without requiring real-time complex decision-making.

The patent uses copying by creating synthetic training datasets that replicate potential attack scenarios. Rather than requiring complex real-world attack data, the system generates copied versions of deceptive signals through data augmentation and simulation. This approach simplifies the training process by providing abundant, diverse training examples without the complexity of collecting and processing real attack data.

The patent applies mechanics substitution by replacing cryptography-based security mechanisms with machine learning-based signal analysis. Instead of relying on encryption to protect sensing signals, the system uses trained ML models to analyze signal characteristics and detect passive sensing attempts. This substitution shifts the security approach from protecting data content to protecting signal integrity through pattern recognition.

The patent applies parameter changes by transforming the security problem from data encryption to signal feature analysis. The system changes the parameters being protected from cryptographic keys and encrypted data to signal characteristics such as temporal patterns, spectral features, and spatial distributions. This parameter transformation enables the detection of passive sensing attempts through anomaly detection in signal parameters rather than through cryptographic verification.