This special section of the IEEE Transactions on Emerging Topics in Computational Intelligence addresses the critical intersection of Privacy, Security, and Computational Intelligence (CI). As the global infrastructure shifts toward the Internet of Things (IoT), cloud computing, and fog computing, the volume of data generated presents both immense opportunities for analysis and significant security risks.
The Dual Challenge of Big Data
The modern "data lifecycle"—encompassing collection, storage, and mining—has been revolutionized by cloud and IoT infrastructure. While CI methods, such as deep learning, are essential for extracting insights from this massive data, they present a double-edged sword:
- Infrastructure Vulnerability: Multi-tenant cloud platforms and ubiquitous IoT access points are inherently prone to privacy leaks and security attacks.
- Algorithmic Gaps: Standard CI algorithms often lack "security-by-design," making them vulnerable to adversarial attacks or unintended data exposure.
This special section bridges the gap between high-performance CI and robust information security through eight peer-reviewed contributions.
Methodological Innovations
The curated papers highlight diverse methodologies designed to secure the data lifecycle:
- Hybrid Feature Matching: To detect image forgeries and partial duplicates, one paper combines "handcrafted" SIFT (Scale Invariant Feature Transform) features with deep CNN features. SIFT provides robustness to rescaling and noise, while CNNs offer high discriminative power.
- Enhanced Stacked LSTM: For malware detection, a deep recurrent neural network replaces random initialization with a pre-training regularization method, significantly improving stability and detection accuracy across Windows, Android, and IoT platforms.
- User-Centric Blockchain (UCB): A framework designed to share "scattered knowledge" across decentralized IoT edge nodes (DINEs) while maintaining auditability and low latency without a central authority.
- Multi-Objective Optimization (MDPW): A method for data placement in cloud workflows that treats privacy as a formal constraint, balancing resource usage and energy costs against the risk of privacy conflicts.
Key Insights and Results
Research within this section demonstrates that purely deep-learning-based approaches often struggle with localized duplicates or adversarial threats. By re-integrating traditional computer vision (like SIFT) or utilizing specialized initialization techniques, researchers achieved superior robustness.
Furthermore, the SSG-MGPM algorithm for graph pattern matching in social networks showed that heuristic approaches can greatly save execution time when network structures change dynamically. The section also includes a landmark survey investigating the last few decades of CI-based information security, identifying current state-of-the-art gaps.
Future Directions
The editors identify several areas for continued evolution:
- Hardware Integration: Implementing privacy concerns directly into hardware instructions, such as specialized neural network processors.
- Consumer Adoption: Overcoming bottlenecks in e-invoicing and other B2C transactions by improving integration into existing consumer behaviors.
- The IoT Arms Race: Moving toward automated, self-healing defense technologies as IoT malware and vulnerabilities become more sophisticated.
Reference: Yuan, Y., Zhang, X., & Tang, J. (2020). Guest Editorial: Special Section on Privacy and Security in Computational Intelligence. IEEE Transactions on Emerging Topics in Computational Intelligence, 4(5), 590-592.
Summary updated