Advancements in Cybersecurity and Hardware Integrity

The recent publications in the field of cybersecurity and hardware integrity reveal a significant shift towards addressing vulnerabilities in critical systems through innovative detection and mitigation strategies. A notable trend is the development of methodologies that ensure the integrity and security of neural networks and hardware components without requiring extensive access or cooperation from the model or hardware owners. This includes leveraging power analysis for black-box neural network integrity checking and employing advanced reverse engineering techniques for hardware IP protection. Additionally, there's a growing emphasis on enhancing the resilience of cyber-physical systems against sophisticated attacks, such as False Data Injection Attacks, through improved detection and data reconstruction strategies. The field is also witnessing advancements in the detection of faults in deep neural networks and the mitigation of adversarial hardware faults in space, highlighting the importance of reliability and security in critical applications. Furthermore, the exploration of machine learning techniques for secure traffic in NoC-based manycores and the development of exact soft analytical side-channel attacks underscore the ongoing efforts to bolster system security against evolving threats.

Noteworthy Papers

  • Michscan: Introduces a novel approach for runtime integrity checking of black-box neural networks using power analysis, demonstrating high accuracy in detecting model integrity violations.
  • CIBPU: Proposes a conflict-invisible secure branch prediction unit that significantly reduces performance overhead while maintaining strong security, marking a notable advancement in secure processor design.
  • ShadowGenes: Presents a signature-based method for machine learning model genealogy, enabling the identification of model architecture and potential security risks with high precision.
  • DEFault: Offers a comprehensive technique for detecting and diagnosing faults in deep neural networks, showcasing improved reliability through hierarchical and explainable classification.
  • Logical Maneuvers: Develops a countermeasure for detecting and mitigating adversarial hardware faults in space, ensuring continuous operation of satellite processors despite permanent faults.

Sources

Michscan: Black-Box Neural Network Integrity Checking at Runtime Through Power Analysis

A Review of Detection, Evolution, and Data Reconstruction Strategies for False Data Injection Attacks in Power Cyber-Physical Systems

CIBPU: A Conflict-Invisible Secure Branch Prediction Unit

ShadowGenes: Leveraging Recurring Patterns within Computational Graphs for Model Genealogy

Application of Machine Learning Techniques for Secure Traffic in NoC-based Manycores

Library-Attack: Reverse Engineering Approach for Evaluating Hardware IP Protection

Improved Detection and Diagnosis of Faults in Deep Neural Networks Using Hierarchical and Explainable Classification

Intelligent Attacks on Cyber-Physical Systems and Critical Infrastructures

Analyzing and Exploiting Branch Mispredictions in Microcode

Real-Time Multi-Modal Subcomponent-Level Measurements for Trustworthy System Monitoring and Malware Detection

Extraction of Secrets from 40nm CMOS Gate Dielectric Breakdown Antifuses by FIB Passive Voltage Contrast

Are We Learning the Right Features?A Framework for Evaluating DL-Based Software Vulnerability Detection Solutions

Exact Soft Analytical Side-Channel Attacks using Tractable Circuits

Logical Maneuvers: Detecting and Mitigating Adversarial Hardware Faults in Space

Built with on top of