Probabilistic Model and Neural Network for Scene Classification in Traffic Surveillance System

Tran, Duong Nguyen-Ngoc; Pham, Long Hoang; Tran, Ha Manh; Ha, Synh Viet-Uyen

doi:10.1007/978-981-10-7512-4_68

Cited by 6 publications

(9 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this paper, we consider X2X backdoor attacks, which refer to a broad family of backdoor attacks with an arbitrary number of source classes, each assigned with an arbitrary target class. Thus, the X2X attack includes many popular attack types such as the "all-to-one" attack, "X-to-one" attack (Shen et al, 2021), "one-to-one" attack (Tran et al, 2018), and "all-to-all" attack (Gu et al, 2019). Unlike other advanced attacks that mostly rely on additional assumptions such as full control of the training process (Zhao et al, 2022;Wang et al, 2022), X2X attacks can be easily launched by poisoning the training set just like classical backdoor attacks.…”

Section: Proceedings Of Thementioning

confidence: 99%

“…Training-phase defenses aim to obtain a backdoor-free classifier from the possibly poisoned training set (Tran et al, 2018;Chen et al, 2018;Xiang et al, 2019;Du et al, 2020;Huang et al, 2022). They can not be deployed at the user end where the classifier is already trained.…”

Section: Related Workmentioning

confidence: 99%

“…It covers many popular attacks with different settings including the "all-to-one" (A2O) attack (Chen et al, 2017), "X-to-one" (X2O) attack (Shen et al, 2021) (a.k.a. a "partial backdoor" (Wang et al, 2019)), "one-to-one" (O2O) attack (Tran et al, 2018), and "all-to-all" (A2A) attack (Gu et al, 2019). The complete taxonomy of X2X backdoor attacks is shown in Fig.…”

Section: Threat Modelmentioning

confidence: 99%

See 2 more Smart Citations

Reverse engineering imperceptible backdoor attacks on deep neural networks for detection and training set cleansing

Xiang

Miller

Kesidis

2021

Computers & Security

View full text Add to dashboard Cite

Section: Proceedings Of Thementioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Threat Modelmentioning

confidence: 99%

See 1 more Smart Citation

Reverse engineering imperceptible backdoor attacks on deep neural networks for detection and training set cleansing

Xiang

Miller

Kesidis

2021

Computers & Security

View full text Add to dashboard Cite

“…The development of the Greedy algorithm can be used by classifying the pattern of extracting data by extracting data, so the analysis of a Greedy algorithm will be clearer, therefore with the addition of the Greedy algorithm, the analysis of each violation will result in a lawsuit in the form of missing and data based on the ticket. can be accounted for legally what articles can be prosecuted in a ticket with the Greedy algorithm, the system will be assisted in analyzing an error committed by motorists on the highway with the Greedy algorithm so all decisions can be taken directly by the Greedy algorithm and results in instantaneous loss [90].…”

Section: Greedy Algorithm Developmentmentioning

confidence: 99%

Application of Greedy Algorithm on Traffic Violation Enforcement

Dewi¹,

Putra²

2021

IJEME

View full text Add to dashboard Cite

The background of this research is how the application of the algorithm in traffic control systems, with the Greedy algorithm, the system will decide what action and what punishment will be given to traffic offenders on the highway, with the Greedy algorithm, the decisions taken will be based on data and facts. based on existing laws, so decisions made based on law and the human side cannot influence the decisions to be taken by the system based on the application of the Greedy algorithm. The research method used in this study uses literature reviews by reading many previous research journals, it will be able to add knowledge and deepen the research we are doing this time, with the literature review method, we will be able to find new problems and can be used as new research because The literature review is very helpful for our research this time. The system that is being used is using CCTV and can determine what decisions and punishments will be given to traffic offenders, through evidence based on images taken by cameras placed at red lights or corners of the capital's highway, the system this has been active effectively but with the implementation of the algorithm will increase. This research will produce a system proposal and be able to find out whether the application of the Greedy algorithm is correct and can help the current system by implementing the algorithm, so the existing system is more perfect. The main contribution of this research is that the use of the Greedy algorithm can help control the traffic system to enforce the law.

show abstract

“…We also evaluate defensive measures against the attack. The most popular backdoor defense detection methods, such as spectral signature detection [30] and activation clustering, [31] struggle to counteract our attack.…”

Section: Introductionmentioning

confidence: 99%

A backdoor attack against quantum neural networks with limited information

Huang

Zhang

2023

Chinese Phys. B

View full text Add to dashboard Cite

Backdoor attacks are emerging security threats to deep neural networks. In these attacks, adversaries manipulate the network by constructing training samples embedded with backdoor triggers. The backdoored model performs as expected on clean test samples but consistently misclassifies samples containing the backdoor trigger as a specific target label. While quantum neural networks (QNNs) have shown promise in surpassing their classical counterparts in certain machine learning tasks, they are also susceptible to backdoor attacks. However, current attacks on QNNs are constrained by the adversary’s understanding of the model structure and specific encoding methods. Given the diversity of encoding methods and model structures in QNNs, the effectiveness of such backdoor attacks remains uncertain. In this paper, we propose an algorithm that leverages dataset-based optimization to initiate backdoor attacks. A malicious adversary can embed backdoor triggers into a QNN model by poisoning only a small portion of the data. The victim QNN maintains high accuracy on clean test samples without the trigger but outputs the target label set by the adversary when predicting samples with the trigger. Furthermore, our proposed attack cannot be easily resisted by existing backdoor detection methods.

show abstract

Probabilistic Model and Neural Network for Scene Classification in Traffic Surveillance System

Cited by 6 publications

References 7 publications

Reverse engineering imperceptible backdoor attacks on deep neural networks for detection and training set cleansing

Reverse engineering imperceptible backdoor attacks on deep neural networks for detection and training set cleansing

Application of Greedy Algorithm on Traffic Violation Enforcement

A backdoor attack against quantum neural networks with limited information

Contact Info

Product

Resources

About