AI2030 Responsible AI Tools Marketplace

Key Takeaways

Explainability can be a practical defense tool: The paper demonstrates that using SHAP (a technique that explains which input features matter most for AI decisions) can help identify and remove harmful manipulations in adversarial attacks, improving classification accuracy by 15-20% across different attack levels.
Current defense methods are inefficient: Existing approaches to protecting AI systems from adversarial attacks suffer from high computational costs and require extensive parameter tuning, suggesting that more practical solutions are needed for real-world deployment in communication networks.
Removing "negative information" improves robustness: By identifying which features have been corrupted by adversarial attacks through explainability analysis and then retraining the model on cleaned data, the system becomes more resistant to multiple types of attacks without requiring complex modifications.
Resource efficiency matters for responsible deployment: The proposed SHAP-based approach not only improves security but also reduces computational resource consumption, which is important for deploying AI systems responsibly in communication infrastructure where efficiency and reliability are critical.
Transparency supports security: Using explainable AI methods to understand why a model makes decisions creates a pathway to both improve security and maintain interpretability—addressing two key concerns in responsible AI development simultaneously.

Abstract

Deep learning (DL) has significantly improved automatic modulation classification (AMC) by leveraging neural networks as the feature extractor.However, as the DL-based AMC becomes increasingly widespread, it is faced with the severe secure issue from various adversarial attacks. Existing defense methods often suffer from the high computational cost, intractable parameter tuning, and insufficient robustness.This paper proposes an eXplainable artificial intelligence (XAI) defense approach, which uncovers the negative information caused by the adversarial attack through measuring the importance of input features based on the SHapley Additive exPlanations (SHAP).By properly removing the negative information in adversarial samples and then fine-tuning(FT) the model, the impact of the attacks on the classification result can be mitigated.Experimental results demonstrate that the proposed SHAP-FT improves the classification performance of the model by 15%-20% under different attack levels,which not only enhances model robustness against various attack levels but also reduces the resource consumption, validating its effectiveness in safeguarding communication networks.

Accepted by WCSP 2024

Key Takeaways

Explainability can be a practical defense tool: The paper demonstrates that using SHAP (a technique that explains which input features matter most for AI decisions) can help identify and remove harmful manipulations in adversarial attacks, improving classification accuracy by 15-20% across different attack levels.
Current defense methods are inefficient: Existing approaches to protecting AI systems from adversarial attacks suffer from high computational costs and require extensive parameter tuning, suggesting that more practical solutions are needed for real-world deployment in communication networks.
Removing "negative information" improves robustness: By identifying which features have been corrupted by adversarial attacks through explainability analysis and then retraining the model on cleaned data, the system becomes more resistant to multiple types of attacks without requiring complex modifications.
Resource efficiency matters for responsible deployment: The proposed SHAP-based approach not only improves security but also reduces computational resource consumption, which is important for deploying AI systems responsibly in communication infrastructure where efficiency and reliability are critical.
Transparency supports security: Using explainable AI methods to understand why a model makes decisions creates a pathway to both improve security and maintain interpretability—addressing two key concerns in responsible AI development simultaneously.

Abstract

Accepted by WCSP 2024

Information Importance-Aware Defense against Adversarial Attack for Automatic Modulation Classification:An XAI-Based Approach

Key Takeaways

Abstract

Paper Info

Responsible AI Topics

Links

Information Importance-Aware Defense against Adversarial Attack for Automatic Modulation Classification:An XAI-Based Approach

Key Takeaways

Abstract

Paper Info

Responsible AI Topics

Links