Machine learning training data poisoning attack defense method

Abstract

The invention belongs to the field of information security, and particularly relates to a machine learning training data poisoning attack defense method, which is used for intelligent security and protection, training data are collected in security and protection video data, and the method comprises the following steps: obtaining prediction value distribution of a clean training data set corresponding to a training data set to be identified; inputting each piece of to-be-identified training data into the trained prediction model to obtain a prediction value; determining and identifying poisoning data based on the predicted value and the predicted value distribution so as to realize defense; generating synthetic data based on the trusted training data of the same type as the training data to be identified; training by adopting an enhanced data set consisting of the synthetic data and the trustworthy training data to obtain a prediction model; wherein the distribution of the enhanced data set is the same as that of the clean training data set, and the prediction model takes the distribution of prediction values output by the enhanced data set as the prediction value distribution. Effective protection can be provided without defining a machine learning algorithm or an attack type, and the problem of limitation on specific attacks in an existing defense method is solved.

Description

technical field [0001] The invention belongs to the field of information security, and more specifically relates to a method for defending machine learning training data from poisoning attacks. Background technique [0002] In recent years, with the development of machine learning, various systems based on machine learning, such as automatic driving systems, face detection systems, and voice recognition systems, have been widely used, especially intelligent security systems. However, various security issues faced by machine learning itself are gradually emerging. [0003] Machine learning refers to continuously learning, identifying features, and modeling through a large amount of training data, and finally an effective system model can be obtained. Recent studies have shown that machine learning is highly susceptible to data poisoning attacks. In this case, an attacker can disrupt the learning process by injecting a small number of malicious samples into the training data...

AI Technical Summary

Problems solved by technology

[0005] The present invention provides a defense method for machine learning training data subjected to poisoning attacks, which is used to solve the limitation problem of only specific attack methods in the existing defense methods for training data used in intelligent security detection model training against poisoning attacks

Images