Device-Specific Signal Timing for Side-Channel Attack Resistance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing countermeasures against side-channel attacks, such as deep learning models, are vulnerable to attacks due to the difficulty in maintaining identical data distributions between training and inference devices, requiring impractical computation to account for the entire trace data.
Innovation Solution
A computer-implemented method modifies the execution of sensitive algorithms by generating signal modification data based on a unique device secret, altering the time evolution of signals with unpredictable countermeasures to ensure that POIs do not align across devices, making it difficult for attackers to derive critical values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If classical countermeasures based on random mask values and shuffling are used, then implementation is simple, but they can be bypassed by deep learning models that learn to recognize and detect them
Solution Approach 1:
The patent changes the fundamental parameter of countermeasure implementation from static random masking to dynamic, device-specific execution timing modifications. By deriving countermeasures from a unique device secret, the system transforms the countermeasure parameters (timing offsets, signal modifications) based on device identity, making them adaptive rather than fixed, thus preventing model generalization.
Solution Approach 2:
The patent segments the countermeasure application by device instance. Each device receives customized countermeasures derived from its unique secret, creating device-specific segmentation of the protection mechanism. This prevents a single trained model from effectively protecting multiple devices, as each device presents a unique countermeasure pattern to the attacker.
2Measurement precision
If deep learning models are trained on traces from a training device, then the model can predict secrets from training data, but the same model cannot be effectively applied to inference devices due to different data distributions
Solution Approach 1:
The device secret acts as an intermediary that mediates between the sensitive algorithm execution and the observable traces. By introducing this secret-derived intermediary layer, the system creates a unique transformation for each device, preventing direct transfer of learning from training to inference devices while maintaining protection effectiveness.
Solution Approach 2:
The patent introduces dynamic, device-specific modifications to the execution timing and signal characteristics based on the device secret. This dynamic adaptation ensures that traces from different devices follow different distributions, making cross-device model transfer ineffective while maintaining security on each individual device.
3Measurement precision
If an attacker trains a model considering the entire trace data, then the model may achieve better accuracy, but it requires unrealistic computation capabilities
Solution Approach 1:
The patent implements excessive action by applying device-specific countermeasures that create sufficient complexity to render full-trace analysis computationally infeasible. The countermeasures introduce enough variability and complexity that attackers must process entire traces with unrealistic computational resources, while the system maintains efficient operation through deterministic countermeasure generation.
Data Source
AI summary
A computer-implemented method of modifying the execution of a sensitive algorithm by a device having a secret S associated with it, wherein execution of the sensitive algorithm comprises using a critical value ki. The computer-implemented method comprises: generating signal modification data based on the secret S, the signal modification data defining one or more countermeasures to be executed during execution of the sensitive algorithm; and executing the sensitive algorithm and one or more countermeasures according to the signal modification data.