InPrivate Digging: Enabling Tree-based Distributed Data Mining with Differential Privacy

Zhao, Lingchen; Ni, Lihao; Hu, Shengshan; Chen, Yaniiao; Zhou, Pan; Xiao, Fu; Wu, Libing

doi:10.1109/infocom.2018.8486352

Cited by 81 publications

(78 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Secondly, it has a rigorous mathematical model, which facilitates quantitative theoretical analysis and proof of privacy levels. In [13], the application of differential privacy in data protection and data mining is demonstrated. Differential privacy is now used in social networks, recommendation systems, network tracking analysis and many other fields.…”

Section: A Data Privacy Protection In Iotmentioning

confidence: 99%

DynaPro: Dynamic Wireless Sensor Network Data Protection Algorithm in IoT via Differential Privacy

Liu

Huang

et al. 2019

IEEE Access

View full text Add to dashboard Cite

With the advent of the era of intelligent Internet of Things (IoT), more and more personal information are collected in the process of deploying IoT in various domains. This brings the problem of data protection in IoT. To address this problem, we propose a data protection algorithm, called DynaPro, based on differential privacy for dynamic wireless sensor networks (WSN), which is the sensor layer of IoT. DynaPro has addressed two issues in data protection of wireless sensor networks. The first issue is the dynamic topology of the network. In order to solve this problem, DynaPro has adopted methods like time window, hierarchical sampling, and graph similarity to process the snapshots of the dynamic network topology. The second issue is addition of noise in differential privacy. To address this issue, DynaPro uses Hierarchical Random Graph (HRG) as middleware to apply differential privacy protection. Instead of adding noise to network data and topology directly, DynaPro adds noise to HRG. In this way, DynaPro can hide the network topology but retain the necessary structural information to support data analysis. Theoretical analysis and experimental results show that DynaPro can preserve the important network features of the original network topology under the premise of the differential privacy protection model.

show abstract

Section: A Data Privacy Protection In Iotmentioning

confidence: 99%

DynaPro: Dynamic Wireless Sensor Network Data Protection Algorithm in IoT via Differential Privacy

Liu

Huang

et al. 2019

IEEE Access

View full text Add to dashboard Cite

show abstract

“…On the other hand, Gradient Boosting Decision Trees (GBDTs) have become very successful in recent years by winning many awards in machine learning and data mining competitions (Chen and Guestrin 2016) as well as their effectiveness in many applications (Richardson, Dominowska, and Ragno 2007;Kim et al 2009;Burges 2010;Li et al 2019b). There have been several recent studies on how to train GBDTs in the federated learning setting (Cheng et al 2019;Liu et al 2019;Zhao et al 2018). For example, SecureBoost (Cheng et al 2019) developed vertical learning with GBDTs.…”

Section: Introductionmentioning

confidence: 99%

“…There have been several studies of GDBT training in the setting of horizontal learning (Liu et al 2019;Zhao et al 2018). However, those approaches are not effective or efficient enough for practical use.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Practical Federated Gradient Boosting Decision Trees

Wen

2020

AAAI

144

View full text Add to dashboard Cite

Gradient Boosting Decision Trees (GBDTs) have become very successful in recent years, with many awards in machine learning and data mining competitions. There have been several recent studies on how to train GBDTs in the federated learning setting. In this paper, we focus on horizontal federated learning, where data samples with the same features are distributed among multiple parties. However, existing studies are not efficient or effective enough for practical use. They suffer either from the inefficiency due to the usage of costly data transformations such as secure sharing and homomorphic encryption, or from the low model accuracy due to differential privacy designs. In this paper, we study a practical federated environment with relaxed privacy constraints. In this environment, a dishonest party might obtain some information about the other parties' data, but it is still impossible for the dishonest party to derive the actual raw data of other parties. Specifically, each party boosts a number of trees by exploiting similarity information based on locality-sensitive hashing. We prove that our framework is secure without exposing the original record to other parties, while the computation overhead in the training process is kept low. Our experimental studies show that, compared with normal training with the local data of each party, our approach can significantly improve the predictive accuracy, and achieve comparable accuracy to the original GBDT with the data from all parties.

show abstract

“…There are two major ways of communications in federated learning: centralized design and distributed design. In the distributed design, communication is performed between two participants, and each participant can directly update global parameters [21]. The data of different participants has the characteristics of independent and identical distribution, which is very suitable for designing efficient training algorithms.…”

Section: Introductionmentioning

confidence: 99%

Heterogeneous Defect Prediction Based on Federated Transfer Learning via Knowledge Distillation

2021

View full text Add to dashboard Cite

Heterogeneous defect prediction (HDP) aims to predict defect-prone software modules in one project using heterogeneous data collected from other projects. There are two characteristics of defect data: data islands, and data privacy. In this paper, we propose a novel Federated Transfer Learning via Knowledge Distillation (FTLKD) approach for HDP, which takes into consideration two characteristics of defect data. Firstly, Shamir sharing technology achieves homomorphic encryption for private data. During subsequent processing and operations, data remains encrypted all the time. Secondly, each participant uses public data to train convolutional neural networks(CNN), the parameters of the pre-trained CNN are transferred to a private model. A small amount of labeled private data fine-tunes the private model. Finally, knowledge distillation realizes the communication between the participants. The average of all softmax output (logits) is used for knowledge distillation to update the private models. Extensive experiments on 9 projects in 3 public databases (NASA, AEEEM and SOFTLAB) show that FTLKD outperforms the related competing methods.

show abstract

InPrivate Digging: Enabling Tree-based Distributed Data Mining with Differential Privacy

Cited by 81 publications

References 16 publications

DynaPro: Dynamic Wireless Sensor Network Data Protection Algorithm in IoT via Differential Privacy

DynaPro: Dynamic Wireless Sensor Network Data Protection Algorithm in IoT via Differential Privacy

Practical Federated Gradient Boosting Decision Trees

Heterogeneous Defect Prediction Based on Federated Transfer Learning via Knowledge Distillation

Contact Info

Product

Resources

About