Malicious Training

Malicious training, a significant threat to machine learning, involves manipulating training data to compromise the integrity and performance of trained models. Current research focuses on developing robust defenses against various attack vectors, including data poisoning (introducing malicious samples) and adversarial examples (subtly altered inputs), employing techniques like temporal analysis of data timestamps, influence graphs to identify malicious samples, and data augmentation to improve model robustness. Understanding and mitigating these attacks is crucial for ensuring the reliability and security of machine learning systems across diverse applications, from malware detection to critical infrastructure.

Papers

October 9, 2024

Hallucinating AI Hijacking Attack: Large Language Models and Malicious Code Recommenders
David Noever, Forrest McKee
Code Repository Code Recommendation Malware Representation Malicious Code Malicious Training

April 19, 2024

Towards Robust Real-Time Hardware-based Mobile Malware Detection using Multiple Instance Learning Formulation
Harshit Kumar, Sudarshan Sharma, Biswadeep Chakraborty, Saibal Mukhopadhyay
Real Time Multiple Instance Learning Android Malware Malware Representation Novel Malware Malicious Training Malware Feature

October 25, 2023

RAEDiff: Denoising Diffusion Probabilistic Models Based Reversible Adversarial Examples Self-Generation and Self-Recovery
Fan Xing, Xiaoyi Zhou, Xuefeng Fan, Zhuo Tian, Yan Zhao
Deep Neural Network Adversarial Example Adversarial Perturbation Denoising Diffusion Probabilistic Model Adversarial Capability Malicious Training

February 7, 2023

Temporal Robustness against Data Poisoning
Wenxiao Wang, Soheil Feizi
Data Poisoning Attack Threat Model Data Poisoning Temporal Robustness Malicious Training

October 18, 2022

Towards Fair Classification against Poisoning Attacks
Han Xu, Xiaorui Liu, Yuxuan Wan, Jiliang Tang
Poisoning Attack Fair Classification Fairness Aware Protected Attribute Fairness Performance Trade Malicious Training

June 7, 2022

Marvolo: Programmatic Data Augmentation for Practical ML-Driven Malware Detection
Michael D. Wong, Edward Raff, James Holt, Ravi Netravali
Malware Detection Malware Datasets Preserving Transformation Based Malware Detection Malicious Training

February 21, 2022

A Tutorial on Adversarial Learning Attacks and Countermeasures
Cato Pauling, Michael Gimson, Muhammed Qaid, Ahmad Kida, Basel Halak
Adversarial Attack Adversarial Learning Tutorial Review State of the Art Defense Rapid Countermeasure Model Stealing Malicious Training

November 29, 2021

A General Framework for Defending Against Backdoor Attacks via Influence Graph
Xiaofei Sun, Jiwei Li, Xiaoya Li, Ziyao Wang, Tianwei Zhang, Han Qiu, Fei Wu, Chun Fan
Backdoor Attack General Framework Pairwise Influence Influence Graph Malicious Training