Quantum fuzzy machine learning adversarial defense model method

A machine learning and quantum technology, applied in quantum computers, computing models, integrated learning, etc., to achieve high robustness, improved robustness, and wide defense applications

CN112488225APending Publication Date: 2021-03-12CHENGDU UNIV OF INFORMATION TECH
0 Cites 3 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Publication Date
2021-03-12

Smart Images

  • Figure 1
    Figure 1
  • Figure 2
    Figure 2
  • Figure 3
    Figure 3
Patent Text Reader

Abstract

The invention discloses a quantum fuzzy machine learning adversarial defense model method. The method comprises the steps of S1, constructing a quantum fuzzy data sample of a legal user; S2, simulating a malicious attacker to construct an attack strategy: adding the constructed disturbance into a quantum fuzzy data sample of a legal user to form a quantum fuzzy countermeasure sample of the malicious attacker; S3, submitting the quantum fuzzy data sample of the legal user and the quantum fuzzy adversarial sample of the malicious attacker to a quantum fuzzy machine learning system for training and learning, and making a correct decision by the quantum fuzzy machine learning system, wherein the quantum fuzzy machine learning system comprises an adversarial defense module, and the adversarialdefense module is an adversarial sample for defending malicious attackers, so that the quantum fuzzy machine learning system makes a correct decision. The model method can effectively resist the attack of a malicious attacker, improve the safety and robustness of a quantum fuzzy machine learning system, and ensure the safe and reliable operation of a quantum fuzzy machine learning algorithm.
Need to check novelty before this filing date? Find Prior Art

Description

technical field

[0001] The invention relates to the fields of quantum machine learning, fuzzy set theory and network confrontation, in particular to a quantum fuzzy machine learning confrontation defense model. Background technique

[0002] In recent years, many research results have been achieved in the field of machine learning, and successful applications have been achieved in many fields, but machine learning also faces many security risks. For example, machine learning systems are easily fooled by adversarial examples, resulting in wrong classifications; users who use online machine learning systems for classification have to disclose their data to the server, which will lead to privacy leaks, and what is worse is the widespread use of machine learning. Use is exacerbating these security risks. At present, many researchers are exploring and studying potential attacks on deep learning and corresponding defense techniques. In the past three years, research on machine le...

Examples

Embodiment 1

[0045] A quantum fuzzy machine learning confrontation defense model method, its flow chart is as follows figure 1 As shown, the details are as follows:

[0046] S1. Constructing a quantum fuzzy data sample of a legitimate user;

[0047] First, according to the quantum fuzzy mathematical management model, fuzzy sets are introduced, and the classic fuzzy data sample set is: D={i ,y i ,μ i (x i )>|x i ∈X},y i ∈{-1,+1},

[0048] Among them, μ i (x i ) for x i Belongs to the fuzzy set {i ,μ A (x i )>|x i ∈X} membership function, each x i =(x i1 ,x i2 ,...,x im ) m eigenvectors are coded into the quantum probability amplitude to form the quantum probability amplitude code, the process can be expressed as:

[0049] Represents a normalized vector;

[0050] Secondly, the quantum fuzzy data sample is prepared, and the quantum fuzzy data sample is expressed as:

[0051] in, x i is the i-th data sample, y i There are only two values ​​(+1 or -1).

[0052] S2. ...

Embodiment 2

[0062] Based on the above embodiment 1, its flow chart is as follows figure 1 and 2 As shown, when the adversarial defense module adopts the first type of defense strategy; that is, the quantum fuzzy machine learning adversarial sample recognizer, it can prevent the samples submitted by malicious attackers and achieve the purpose of defense.

[0063] Among them, the defense method of quantum fuzzy machine learning against sample identifier is as follows:

[0064] The quantum fuzzy machine learning adversarial example recognizer uses a binary classification method for training and testing, and identifies quantum fuzzy data samples submitted by legitimate users to make correct decisions; identify quantum fuzzy adversarial examples submitted by malicious attackers Block the attack of this sample.

[0065] Among them, the two classification methods include:

[0066] Divide the quantum fuzzy data samples into a test set of quantum fuzzy data samples and the training set of ...

Embodiment 3

[0070] Based on the above embodiment 1, its flow chart is as follows figure 1 and 3 As shown, when the adversarial defense module adopts the second type of defense strategy, the second type of defense strategy is to reconstruct the input quantum fuzzy mixed samples (including quantum fuzzy data samples and quantum fuzzy confrontation samples), so that the quantum fuzzy machine learning system can do Make the right decision to achieve the goal of defense.

[0071] Quantum fuzzy machine learning defenses against sample correctors include:

[0072] Input the quantum fuzzy mixed sample, the quantum fuzzy mixed sample is the sum of the legitimate user's quantum fuzzy data sample and the malicious attacker's quantum fuzzy confrontation sample;

[0073] The quantum fuzzy mixed samples are corrected by the reconstruction method, so that the quantum fuzzy machine learning system can make correct decisions.

[0074] Among them, the reconstruction method is to use the self-encoding me...