SlideShare a Scribd company logo
1 of 17
Frequency-based Constraint Relaxation
for Private Query Processing in Cloud Databases
Junpei Kawamoto (Kyushu University, Japan)
Patricia L. Gillett (École Polytechnique de Montréal)
Cloud services and Privacy
• Cloud services as cloud databases.
• Sometimes people want to keep what they request to databases
secret.
May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 2
Find restaurants near by current location x.
Location based
services
Want to read article x
Forum sites
Private query processing
• Methodologies to obtain data w/o exposing queries.
• Several protocols such as cPIR† & bbPIR†† are introduced.
May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 3
Cloud Database
Current location x
query
res.
Encode x
Compute query results w/o
decoding queries
†Kushilevitz, E. and Ostrovsky, R.: Replication Is Not Needed: Single Database, Computationally-
Private Information Retrieval, Proc. of the 38th Annual Symposium on Foundations of Computer
Science, pp. 364-373, 1997.
E.g. Location Based Services
††Wang, S., Agrawal, D., and Abbadi, A.: Generalizing PIR for Practical Private Retrieval of Public
Data, Proc. of the 24th Annual IFIP WG 11.3 Working Conference on Data and Applications Security
and Privacy, pp. 1-16, 2010.
Three ideas of private queries
• We introduce three ideas for our discussion;
• Search intention: what users hope to obtain from cloud services,
• Query: request users send to servers to obtain data,
• Handled set: data set which servers must check to compute the results.
May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 4
Location x
Want information associated with
location x (search intention)
query
res.
Cloud DB
Must check these items to
compute the result(Handled set)
Handled set
for x
Existing private querying protocols
• Most existing protocols impose two constraints:
1. Queries are encoded in such a way that servers can handle query
processes but cannot actually decode queries;
2. Servers are made to check all data in the databases when computing
any query result.
• (2) means servers cannot distinguish any data in the DB.
• Servers spend O(n) computational cost executing each query, where
the database has n entries.
May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 5
Cloud DB
check all data any time
query
• There are cases in which we do not need to retrieve the
entire database to sufficiently obscure search intentions.
• It may be enough to hide where I am in downtown.
• What area is enough to ensure our privacy?
• E.g.
• Servers may guess the search intention is x with high probability.
• We should consider frequency of search intentions.
Constraint Relaxation
May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 6
shop
x
shop
a
shop
b
shop
c
Popular place Unpopular places
Specified by a handled set
a handled set
Database model & frequency
• Database model
• Database D consists of n items: D = {t1, t2, …, tn}.
• Users request x-th item tx (The search intention is x).
• Handled set H(x)
• The item set servers must check to compute the results of a given
query associated with search intention x.
• Frequency of search intentions
• Freq(x) denotes the frequency of search intention x.
• We assume Freq(x) is normalized.
May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 7
Want tx (search intention)
query
res.
DB
H(x)
Must check these items to
compute the result(Handled set)
Definition: Query Risk
• A measure of exposure risk for private queries.
• The query risk of search intention x and handled set H(x) is
• Conditional probability that the search
intention is x given that we know the
handled set H used.
• The more frequent x is,
the higher the risk becomes.
May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 8


)(
)(Freq
)(Freq
))(|(Risk
xHy
y
x
xHx
frequenc
y
E.g.
• Risk(3|{1,2,3,4,5}) = 3/9
• Risk(1|{1,2,3,4,5}) = 1/9
• Risk(1|{1,2}) = 1/3
Definition: Privateness
• The maximum risk of complete protocols:
• We assume query risks should be less than or equals to
the maximum risk of complete protocols:
• Complete protocols are widely accepted that is the max. risk is, too.
• Our approach will also be considered private if it satisfies this condition.
May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 9
))(|(Riskmax))(|(Risk completerelaxed yHyxHx
Dy

))(|(Riskmax complete xHx
Dx
i.e. w/o relaxation
Our handled set Handled set for existing protocols
Definition: Query processing cost
• Query processing costs on servers.
• We evaluate them by the size of handled sets.
• The cost of search intention x is .
May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 10
|)(|)(Cost xHx 
query
res.
DB
H(x)
H(x)
query
res.
DB
vs.
The problem
• Find handled set H such that, for search intention x:
1. ,
2. ,
3. minimize Cost(x),
4. if multiple solutions have equal cost, chose the one maximizes
• We solve this problem by Dynamic Programming.
• O(n) algorithm but details are in our paper.
• We also extend these problem and algorithm to
• range queries in 1D data,
• exact match queries in 2D data.
May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 11
x Î H(x)
)|(Riskmax))(|(Risk DyxHx Dy
  )(
)(FreqxHy
y
The protocol
• Our protocol employs some existing protocol (PIR, etc.).
• We assume the frequencies are public information.
• User: whose search intension is x,
1. Compute optimized handled set H(x) using query frequencies.
2. Compute a private query for x assuming a DB has only items in H(x).
3. Send the query and H(x) to the cloud server.
• The cloud server receiving the query,
1. Consider a sub-database consists of items in H(x).
2. Process the received query and return the result to the user.
May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 12
query
res. H(x) : sub-DB
Cost: O(|H(x)|)
• Dataset
• Query logs from †.
• Sampled 100,000 songs and 1,800,145 queries.
Evaluation
May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 13
†http://www.dtic.upf.edu/~ocelma/MusicRecommendationDataset/lastfm-1K.html
Frequency of search intention x.
x-axis: search intentions x
y-axis: # of times users requested
item tx
Evaluation
• Comparison of query risks.
• In most cases, risks are bigger than those of the complete protocols.
• Do not exceed the maximum risks of the complete protocol.
May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 14
relaxed (avg.) is computed by


Dx
xHxx ))(|(Risk)(Freq
min. max. avg.
comp. 5.6×10−7 3.9×10−3 1.0×10−5
relaxed 3.4×10−3 3.9×10−3 3.5×10−3
Query risk of search intention x.
Evaluation
• Comparison of query costs.
• Our relaxation methodologies reduce costs in most cases.
• The average cost is 6.5% that of complete protocols.
May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 15
relaxed (avg.) is computed by
Freq(x)´Cost(x)
xÎD
å
min. max. avg.
comp. 100000 100000 100000
relaxed 2 100000 6417
Query cost of search intention x.
Conclusion
• We introduced a frequency-based constraint relaxation
methodology for private queries.
• We relaxed constraint (2) of the complete protocols so that
only a subset of the database is retrieved for each query.
• We evaluated our proposal using a real dataset from .
• Our protocol can reduce computational costs in servers in most cases,
• The risk of a query being exposed is not bigger than the maximum risk
in complete protocols.
May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 16
Servers must check all data in the databases
when computing any query result.
Acknowledgement
• This work is partly supported by
• The Nakajima Foundation,
• Artificial Intelligence Research Promotion Foundation,
• Grant-in-Aid for Young Scientists (B) (26730065), Japan Society for
the Promotion of Science (JSPS).
May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 17
Thank you for your attention!

More Related Content

What's hot

Hands on Mahout!
Hands on Mahout!Hands on Mahout!
Hands on Mahout!OSCON Byrum
 
Using the Open Science Data Cloud for Data Science Research
Using the Open Science Data Cloud for Data Science ResearchUsing the Open Science Data Cloud for Data Science Research
Using the Open Science Data Cloud for Data Science ResearchRobert Grossman
 
Orchestrating the Intelligent Web with Apache Mahout
Orchestrating the Intelligent Web with Apache MahoutOrchestrating the Intelligent Web with Apache Mahout
Orchestrating the Intelligent Web with Apache Mahoutaneeshabakharia
 
Increasing Security Awareness in Enterprise Using Automated Feature Extractio...
Increasing Security Awareness in Enterprise Using Automated Feature Extractio...Increasing Security Awareness in Enterprise Using Automated Feature Extractio...
Increasing Security Awareness in Enterprise Using Automated Feature Extractio...Burman Noviansyah
 
Information Flow and Search in Unstructured Keyword based Social Networks
Information Flow and Search in Unstructured Keyword based Social NetworksInformation Flow and Search in Unstructured Keyword based Social Networks
Information Flow and Search in Unstructured Keyword based Social NetworksPrantik Bhattacharyya
 
What Are Science Clouds?
What Are Science Clouds?What Are Science Clouds?
What Are Science Clouds?Robert Grossman
 
DSTree: A Tree Structure for the Mining of Frequent Sets from Data Streams
DSTree: A Tree Structure for the Mining of Frequent Sets from Data StreamsDSTree: A Tree Structure for the Mining of Frequent Sets from Data Streams
DSTree: A Tree Structure for the Mining of Frequent Sets from Data StreamsAllenWu
 
Keeping Linked Open Data Caches Up-to-date by Predicting the Life-time of RDF...
Keeping Linked Open Data Caches Up-to-date by Predicting the Life-time of RDF...Keeping Linked Open Data Caches Up-to-date by Predicting the Life-time of RDF...
Keeping Linked Open Data Caches Up-to-date by Predicting the Life-time of RDF...MOVING Project
 
HGrid A Data Model for Large Geospatial Data Sets in HBase
HGrid A Data Model for Large Geospatial Data Sets in HBaseHGrid A Data Model for Large Geospatial Data Sets in HBase
HGrid A Data Model for Large Geospatial Data Sets in HBaseDan Han
 
Probablistic information retrieval
Probablistic information retrievalProbablistic information retrieval
Probablistic information retrievalNisha Arankandath
 
Using High Dimensional Representation of Words (CBOW) to Find Domain Based Co...
Using High Dimensional Representation of Words (CBOW) to Find Domain Based Co...Using High Dimensional Representation of Words (CBOW) to Find Domain Based Co...
Using High Dimensional Representation of Words (CBOW) to Find Domain Based Co...HPCC Systems
 
A Survey on Approaches for Frequent Item Set Mining on Apache Hadoop
A Survey on Approaches for Frequent Item Set Mining on Apache HadoopA Survey on Approaches for Frequent Item Set Mining on Apache Hadoop
A Survey on Approaches for Frequent Item Set Mining on Apache HadoopIJTET Journal
 
Presented by Anu Mattatholi
Presented by Anu MattatholiPresented by Anu Mattatholi
Presented by Anu MattatholiAnu Mattatholi
 
How To Analyze Geolocation Data with Hive and Hadoop
How To Analyze Geolocation Data with Hive and HadoopHow To Analyze Geolocation Data with Hive and Hadoop
How To Analyze Geolocation Data with Hive and HadoopHortonworks
 

What's hot (17)

Hands on Mahout!
Hands on Mahout!Hands on Mahout!
Hands on Mahout!
 
Using the Open Science Data Cloud for Data Science Research
Using the Open Science Data Cloud for Data Science ResearchUsing the Open Science Data Cloud for Data Science Research
Using the Open Science Data Cloud for Data Science Research
 
Orchestrating the Intelligent Web with Apache Mahout
Orchestrating the Intelligent Web with Apache MahoutOrchestrating the Intelligent Web with Apache Mahout
Orchestrating the Intelligent Web with Apache Mahout
 
Increasing Security Awareness in Enterprise Using Automated Feature Extractio...
Increasing Security Awareness in Enterprise Using Automated Feature Extractio...Increasing Security Awareness in Enterprise Using Automated Feature Extractio...
Increasing Security Awareness in Enterprise Using Automated Feature Extractio...
 
Information Flow and Search in Unstructured Keyword based Social Networks
Information Flow and Search in Unstructured Keyword based Social NetworksInformation Flow and Search in Unstructured Keyword based Social Networks
Information Flow and Search in Unstructured Keyword based Social Networks
 
Csi
CsiCsi
Csi
 
What Are Science Clouds?
What Are Science Clouds?What Are Science Clouds?
What Are Science Clouds?
 
Jj3616251628
Jj3616251628Jj3616251628
Jj3616251628
 
DSTree: A Tree Structure for the Mining of Frequent Sets from Data Streams
DSTree: A Tree Structure for the Mining of Frequent Sets from Data StreamsDSTree: A Tree Structure for the Mining of Frequent Sets from Data Streams
DSTree: A Tree Structure for the Mining of Frequent Sets from Data Streams
 
Keeping Linked Open Data Caches Up-to-date by Predicting the Life-time of RDF...
Keeping Linked Open Data Caches Up-to-date by Predicting the Life-time of RDF...Keeping Linked Open Data Caches Up-to-date by Predicting the Life-time of RDF...
Keeping Linked Open Data Caches Up-to-date by Predicting the Life-time of RDF...
 
HGrid A Data Model for Large Geospatial Data Sets in HBase
HGrid A Data Model for Large Geospatial Data Sets in HBaseHGrid A Data Model for Large Geospatial Data Sets in HBase
HGrid A Data Model for Large Geospatial Data Sets in HBase
 
Probablistic information retrieval
Probablistic information retrievalProbablistic information retrieval
Probablistic information retrieval
 
Using High Dimensional Representation of Words (CBOW) to Find Domain Based Co...
Using High Dimensional Representation of Words (CBOW) to Find Domain Based Co...Using High Dimensional Representation of Words (CBOW) to Find Domain Based Co...
Using High Dimensional Representation of Words (CBOW) to Find Domain Based Co...
 
A Survey on Approaches for Frequent Item Set Mining on Apache Hadoop
A Survey on Approaches for Frequent Item Set Mining on Apache HadoopA Survey on Approaches for Frequent Item Set Mining on Apache Hadoop
A Survey on Approaches for Frequent Item Set Mining on Apache Hadoop
 
Presented by Anu Mattatholi
Presented by Anu MattatholiPresented by Anu Mattatholi
Presented by Anu Mattatholi
 
How To Analyze Geolocation Data with Hive and Hadoop
How To Analyze Geolocation Data with Hive and HadoopHow To Analyze Geolocation Data with Hive and Hadoop
How To Analyze Geolocation Data with Hive and Hadoop
 
CSCC-X2007
CSCC-X2007CSCC-X2007
CSCC-X2007
 

Similar to Frequency-based Constraint Relaxation for Private Cloud Database Queries

Approximation Data Structures for Streaming Applications
Approximation Data Structures for Streaming ApplicationsApproximation Data Structures for Streaming Applications
Approximation Data Structures for Streaming ApplicationsDebasish Ghosh
 
Information Retrieval Dynamic Time Warping - Interspeech 2013 presentation
Information Retrieval Dynamic Time Warping - Interspeech 2013 presentationInformation Retrieval Dynamic Time Warping - Interspeech 2013 presentation
Information Retrieval Dynamic Time Warping - Interspeech 2013 presentationXavier Anguera
 
Outlier and fraud detection using Hadoop
Outlier and fraud detection using HadoopOutlier and fraud detection using Hadoop
Outlier and fraud detection using HadoopPranab Ghosh
 
Mining high speed data streams: Hoeffding and VFDT
Mining high speed data streams: Hoeffding and VFDTMining high speed data streams: Hoeffding and VFDT
Mining high speed data streams: Hoeffding and VFDTDavide Gallitelli
 
Wastian, Brunmeir - Data Analyses in Industrial Applications: From Predictive...
Wastian, Brunmeir - Data Analyses in Industrial Applications: From Predictive...Wastian, Brunmeir - Data Analyses in Industrial Applications: From Predictive...
Wastian, Brunmeir - Data Analyses in Industrial Applications: From Predictive...Vienna Data Science Group
 
InfluxEnterprise Architectural Patterns by Dean Sheehan, Senior Director, Pre...
InfluxEnterprise Architectural Patterns by Dean Sheehan, Senior Director, Pre...InfluxEnterprise Architectural Patterns by Dean Sheehan, Senior Director, Pre...
InfluxEnterprise Architectural Patterns by Dean Sheehan, Senior Director, Pre...InfluxData
 
"Quantum clustering - physics inspired clustering algorithm", Sigalit Bechler...
"Quantum clustering - physics inspired clustering algorithm", Sigalit Bechler..."Quantum clustering - physics inspired clustering algorithm", Sigalit Bechler...
"Quantum clustering - physics inspired clustering algorithm", Sigalit Bechler...Dataconomy Media
 
Apache Drill: An Active, Ad-hoc Query System for large-scale Data Sets
Apache Drill: An Active, Ad-hoc Query System for large-scale Data SetsApache Drill: An Active, Ad-hoc Query System for large-scale Data Sets
Apache Drill: An Active, Ad-hoc Query System for large-scale Data SetsMapR Technologies
 
Hyperoptimized Machine Learning and Deep Learning Methods For Geospatial and ...
Hyperoptimized Machine Learning and Deep Learning Methods For Geospatial and ...Hyperoptimized Machine Learning and Deep Learning Methods For Geospatial and ...
Hyperoptimized Machine Learning and Deep Learning Methods For Geospatial and ...Neelabha Pant
 
"Quantum Clustering - Physics Inspired Clustering Algorithm", Sigalit Bechler...
"Quantum Clustering - Physics Inspired Clustering Algorithm", Sigalit Bechler..."Quantum Clustering - Physics Inspired Clustering Algorithm", Sigalit Bechler...
"Quantum Clustering - Physics Inspired Clustering Algorithm", Sigalit Bechler...Dataconomy Media
 
Big Data Technologies - Hadoop
Big Data Technologies - HadoopBig Data Technologies - Hadoop
Big Data Technologies - HadoopTalentica Software
 
Шардинг в MongoDB, Henrik Ingo (MongoDB)
Шардинг в MongoDB, Henrik Ingo (MongoDB)Шардинг в MongoDB, Henrik Ingo (MongoDB)
Шардинг в MongoDB, Henrik Ingo (MongoDB)Ontico
 
A Production Quality Sketching Library for the Analysis of Big Data
A Production Quality Sketching Library for the Analysis of Big DataA Production Quality Sketching Library for the Analysis of Big Data
A Production Quality Sketching Library for the Analysis of Big DataDatabricks
 
Differential privacy (개인정보 차등보호)
Differential privacy (개인정보 차등보호)Differential privacy (개인정보 차등보호)
Differential privacy (개인정보 차등보호)Young-Geun Choi
 
InfluxEnterprise Architecture Patterns by Tim Hall & Sam Dillard
InfluxEnterprise Architecture Patterns by Tim Hall & Sam DillardInfluxEnterprise Architecture Patterns by Tim Hall & Sam Dillard
InfluxEnterprise Architecture Patterns by Tim Hall & Sam DillardInfluxData
 
What is a Data Commons and Why Should You Care?
What is a Data Commons and Why Should You Care? What is a Data Commons and Why Should You Care?
What is a Data Commons and Why Should You Care? Robert Grossman
 
State of Florida Neo4j Graph Briefing - Cyber IAM
State of Florida Neo4j Graph Briefing - Cyber IAMState of Florida Neo4j Graph Briefing - Cyber IAM
State of Florida Neo4j Graph Briefing - Cyber IAMNeo4j
 
Azure 機器學習 - 使用Python, R, Spark, CNTK 深度學習
Azure 機器學習 - 使用Python, R, Spark, CNTK 深度學習 Azure 機器學習 - 使用Python, R, Spark, CNTK 深度學習
Azure 機器學習 - 使用Python, R, Spark, CNTK 深度學習 Herman Wu
 

Similar to Frequency-based Constraint Relaxation for Private Cloud Database Queries (20)

Approximation Data Structures for Streaming Applications
Approximation Data Structures for Streaming ApplicationsApproximation Data Structures for Streaming Applications
Approximation Data Structures for Streaming Applications
 
Information Retrieval Dynamic Time Warping - Interspeech 2013 presentation
Information Retrieval Dynamic Time Warping - Interspeech 2013 presentationInformation Retrieval Dynamic Time Warping - Interspeech 2013 presentation
Information Retrieval Dynamic Time Warping - Interspeech 2013 presentation
 
Outlier and fraud detection using Hadoop
Outlier and fraud detection using HadoopOutlier and fraud detection using Hadoop
Outlier and fraud detection using Hadoop
 
05 k-means clustering
05 k-means clustering05 k-means clustering
05 k-means clustering
 
Mining high speed data streams: Hoeffding and VFDT
Mining high speed data streams: Hoeffding and VFDTMining high speed data streams: Hoeffding and VFDT
Mining high speed data streams: Hoeffding and VFDT
 
Wastian, Brunmeir - Data Analyses in Industrial Applications: From Predictive...
Wastian, Brunmeir - Data Analyses in Industrial Applications: From Predictive...Wastian, Brunmeir - Data Analyses in Industrial Applications: From Predictive...
Wastian, Brunmeir - Data Analyses in Industrial Applications: From Predictive...
 
InfluxEnterprise Architectural Patterns by Dean Sheehan, Senior Director, Pre...
InfluxEnterprise Architectural Patterns by Dean Sheehan, Senior Director, Pre...InfluxEnterprise Architectural Patterns by Dean Sheehan, Senior Director, Pre...
InfluxEnterprise Architectural Patterns by Dean Sheehan, Senior Director, Pre...
 
"Quantum clustering - physics inspired clustering algorithm", Sigalit Bechler...
"Quantum clustering - physics inspired clustering algorithm", Sigalit Bechler..."Quantum clustering - physics inspired clustering algorithm", Sigalit Bechler...
"Quantum clustering - physics inspired clustering algorithm", Sigalit Bechler...
 
Apache Drill: An Active, Ad-hoc Query System for large-scale Data Sets
Apache Drill: An Active, Ad-hoc Query System for large-scale Data SetsApache Drill: An Active, Ad-hoc Query System for large-scale Data Sets
Apache Drill: An Active, Ad-hoc Query System for large-scale Data Sets
 
Hyperoptimized Machine Learning and Deep Learning Methods For Geospatial and ...
Hyperoptimized Machine Learning and Deep Learning Methods For Geospatial and ...Hyperoptimized Machine Learning and Deep Learning Methods For Geospatial and ...
Hyperoptimized Machine Learning and Deep Learning Methods For Geospatial and ...
 
"Quantum Clustering - Physics Inspired Clustering Algorithm", Sigalit Bechler...
"Quantum Clustering - Physics Inspired Clustering Algorithm", Sigalit Bechler..."Quantum Clustering - Physics Inspired Clustering Algorithm", Sigalit Bechler...
"Quantum Clustering - Physics Inspired Clustering Algorithm", Sigalit Bechler...
 
Big Data Technologies - Hadoop
Big Data Technologies - HadoopBig Data Technologies - Hadoop
Big Data Technologies - Hadoop
 
Watson Computer
Watson ComputerWatson Computer
Watson Computer
 
Шардинг в MongoDB, Henrik Ingo (MongoDB)
Шардинг в MongoDB, Henrik Ingo (MongoDB)Шардинг в MongoDB, Henrik Ingo (MongoDB)
Шардинг в MongoDB, Henrik Ingo (MongoDB)
 
A Production Quality Sketching Library for the Analysis of Big Data
A Production Quality Sketching Library for the Analysis of Big DataA Production Quality Sketching Library for the Analysis of Big Data
A Production Quality Sketching Library for the Analysis of Big Data
 
Differential privacy (개인정보 차등보호)
Differential privacy (개인정보 차등보호)Differential privacy (개인정보 차등보호)
Differential privacy (개인정보 차등보호)
 
InfluxEnterprise Architecture Patterns by Tim Hall & Sam Dillard
InfluxEnterprise Architecture Patterns by Tim Hall & Sam DillardInfluxEnterprise Architecture Patterns by Tim Hall & Sam Dillard
InfluxEnterprise Architecture Patterns by Tim Hall & Sam Dillard
 
What is a Data Commons and Why Should You Care?
What is a Data Commons and Why Should You Care? What is a Data Commons and Why Should You Care?
What is a Data Commons and Why Should You Care?
 
State of Florida Neo4j Graph Briefing - Cyber IAM
State of Florida Neo4j Graph Briefing - Cyber IAMState of Florida Neo4j Graph Briefing - Cyber IAM
State of Florida Neo4j Graph Briefing - Cyber IAM
 
Azure 機器學習 - 使用Python, R, Spark, CNTK 深度學習
Azure 機器學習 - 使用Python, R, Spark, CNTK 深度學習 Azure 機器學習 - 使用Python, R, Spark, CNTK 深度學習
Azure 機器學習 - 使用Python, R, Spark, CNTK 深度學習
 

More from Junpei Kawamoto

レビューサイトにおける不均質性を考慮した特異なレビュアー発⾒とレビューサマリの推測
レビューサイトにおける不均質性を考慮した特異なレビュアー発⾒とレビューサマリの推測レビューサイトにおける不均質性を考慮した特異なレビュアー発⾒とレビューサマリの推測
レビューサイトにおける不均質性を考慮した特異なレビュアー発⾒とレビューサマリの推測Junpei Kawamoto
 
初期レビューを用いた長期間評価推定􏰀
初期レビューを用いた長期間評価推定􏰀初期レビューを用いた長期間評価推定􏰀
初期レビューを用いた長期間評価推定􏰀Junpei Kawamoto
 
Securing Social Information from Query Analysis in Outsourced Databases
Securing Social Information from Query Analysis in Outsourced DatabasesSecuring Social Information from Query Analysis in Outsourced Databases
Securing Social Information from Query Analysis in Outsourced DatabasesJunpei Kawamoto
 
クエリログとナビゲーション履歴から探索意図抽出による協調探索支援
クエリログとナビゲーション履歴から探索意図抽出による協調探索支援クエリログとナビゲーション履歴から探索意図抽出による協調探索支援
クエリログとナビゲーション履歴から探索意図抽出による協調探索支援Junpei Kawamoto
 
Privacy for Continual Data Publishing
Privacy for Continual Data PublishingPrivacy for Continual Data Publishing
Privacy for Continual Data PublishingJunpei Kawamoto
 
暗号化ベクトルデータベースのための索引構造
暗号化ベクトルデータベースのための索引構造暗号化ベクトルデータベースのための索引構造
暗号化ベクトルデータベースのための索引構造Junpei Kawamoto
 
暗号化データベースモデルにおける問合せの関連情報を秘匿する範囲検索
暗号化データベースモデルにおける問合せの関連情報を秘匿する範囲検索暗号化データベースモデルにおける問合せの関連情報を秘匿する範囲検索
暗号化データベースモデルにおける問合せの関連情報を秘匿する範囲検索Junpei Kawamoto
 
マルコフ過程を用いた位置情報継続開示のためのアドバーザリアルプライバシ
マルコフ過程を用いた位置情報継続開示のためのアドバーザリアルプライバシマルコフ過程を用いた位置情報継続開示のためのアドバーザリアルプライバシ
マルコフ過程を用いた位置情報継続開示のためのアドバーザリアルプライバシJunpei Kawamoto
 
データ共有型WEBアプリケーションにおけるサーバ暗号化
データ共有型WEBアプリケーションにおけるサーバ暗号化データ共有型WEBアプリケーションにおけるサーバ暗号化
データ共有型WEBアプリケーションにおけるサーバ暗号化Junpei Kawamoto
 
マルコフモデルを仮定した位置情報開示のためのアドバーザリアルプライバシ
マルコフモデルを仮定した位置情報開示のためのアドバーザリアルプライバシマルコフモデルを仮定した位置情報開示のためのアドバーザリアルプライバシ
マルコフモデルを仮定した位置情報開示のためのアドバーザリアルプライバシJunpei Kawamoto
 
プライベート問合せにおける問合せ頻度を用いた制約緩和手法
プライベート問合せにおける問合せ頻度を用いた制約緩和手法プライベート問合せにおける問合せ頻度を用いた制約緩和手法
プライベート問合せにおける問合せ頻度を用いた制約緩和手法Junpei Kawamoto
 
プライバシを考慮した移動系列情報解析のための安全性の提案
プライバシを考慮した移動系列情報解析のための安全性の提案プライバシを考慮した移動系列情報解析のための安全性の提案
プライバシを考慮した移動系列情報解析のための安全性の提案Junpei Kawamoto
 
A Locality Sensitive Hashing Filter for Encrypted Vector Databases
A Locality Sensitive Hashing Filter for Encrypted Vector DatabasesA Locality Sensitive Hashing Filter for Encrypted Vector Databases
A Locality Sensitive Hashing Filter for Encrypted Vector DatabasesJunpei Kawamoto
 
位置情報解析のためのプライバシ保護手法
位置情報解析のためのプライバシ保護手法位置情報解析のためのプライバシ保護手法
位置情報解析のためのプライバシ保護手法Junpei Kawamoto
 
Sponsored Search Markets (from Networks, Crowds, and Markets: Reasoning About...
Sponsored Search Markets (from Networks, Crowds, and Markets: Reasoning About...Sponsored Search Markets (from Networks, Crowds, and Markets: Reasoning About...
Sponsored Search Markets (from Networks, Crowds, and Markets: Reasoning About...Junpei Kawamoto
 
Private Range Query by Perturbation and Matrix Based Encryption
Private Range Query by Perturbation and Matrix Based EncryptionPrivate Range Query by Perturbation and Matrix Based Encryption
Private Range Query by Perturbation and Matrix Based EncryptionJunpei Kawamoto
 
暗号化データベースモデルにおける関係情報推定を防ぐ索引手法
暗号化データベースモデルにおける関係情報推定を防ぐ索引手法暗号化データベースモデルにおける関係情報推定を防ぐ索引手法
暗号化データベースモデルにおける関係情報推定を防ぐ索引手法Junpei Kawamoto
 
VLDB09勉強会 Session27 Privacy2
VLDB09勉強会 Session27 Privacy2VLDB09勉強会 Session27 Privacy2
VLDB09勉強会 Session27 Privacy2Junpei Kawamoto
 
Reducing Data Decryption Cost by Broadcast Encryption and Account Assignment ...
Reducing Data Decryption Cost by Broadcast Encryption and Account Assignment ...Reducing Data Decryption Cost by Broadcast Encryption and Account Assignment ...
Reducing Data Decryption Cost by Broadcast Encryption and Account Assignment ...Junpei Kawamoto
 
Security of Social Information from Query Analysis in DaaS
Security of Social Information from Query Analysis in DaaSSecurity of Social Information from Query Analysis in DaaS
Security of Social Information from Query Analysis in DaaSJunpei Kawamoto
 

More from Junpei Kawamoto (20)

レビューサイトにおける不均質性を考慮した特異なレビュアー発⾒とレビューサマリの推測
レビューサイトにおける不均質性を考慮した特異なレビュアー発⾒とレビューサマリの推測レビューサイトにおける不均質性を考慮した特異なレビュアー発⾒とレビューサマリの推測
レビューサイトにおける不均質性を考慮した特異なレビュアー発⾒とレビューサマリの推測
 
初期レビューを用いた長期間評価推定􏰀
初期レビューを用いた長期間評価推定􏰀初期レビューを用いた長期間評価推定􏰀
初期レビューを用いた長期間評価推定􏰀
 
Securing Social Information from Query Analysis in Outsourced Databases
Securing Social Information from Query Analysis in Outsourced DatabasesSecuring Social Information from Query Analysis in Outsourced Databases
Securing Social Information from Query Analysis in Outsourced Databases
 
クエリログとナビゲーション履歴から探索意図抽出による協調探索支援
クエリログとナビゲーション履歴から探索意図抽出による協調探索支援クエリログとナビゲーション履歴から探索意図抽出による協調探索支援
クエリログとナビゲーション履歴から探索意図抽出による協調探索支援
 
Privacy for Continual Data Publishing
Privacy for Continual Data PublishingPrivacy for Continual Data Publishing
Privacy for Continual Data Publishing
 
暗号化ベクトルデータベースのための索引構造
暗号化ベクトルデータベースのための索引構造暗号化ベクトルデータベースのための索引構造
暗号化ベクトルデータベースのための索引構造
 
暗号化データベースモデルにおける問合せの関連情報を秘匿する範囲検索
暗号化データベースモデルにおける問合せの関連情報を秘匿する範囲検索暗号化データベースモデルにおける問合せの関連情報を秘匿する範囲検索
暗号化データベースモデルにおける問合せの関連情報を秘匿する範囲検索
 
マルコフ過程を用いた位置情報継続開示のためのアドバーザリアルプライバシ
マルコフ過程を用いた位置情報継続開示のためのアドバーザリアルプライバシマルコフ過程を用いた位置情報継続開示のためのアドバーザリアルプライバシ
マルコフ過程を用いた位置情報継続開示のためのアドバーザリアルプライバシ
 
データ共有型WEBアプリケーションにおけるサーバ暗号化
データ共有型WEBアプリケーションにおけるサーバ暗号化データ共有型WEBアプリケーションにおけるサーバ暗号化
データ共有型WEBアプリケーションにおけるサーバ暗号化
 
マルコフモデルを仮定した位置情報開示のためのアドバーザリアルプライバシ
マルコフモデルを仮定した位置情報開示のためのアドバーザリアルプライバシマルコフモデルを仮定した位置情報開示のためのアドバーザリアルプライバシ
マルコフモデルを仮定した位置情報開示のためのアドバーザリアルプライバシ
 
プライベート問合せにおける問合せ頻度を用いた制約緩和手法
プライベート問合せにおける問合せ頻度を用いた制約緩和手法プライベート問合せにおける問合せ頻度を用いた制約緩和手法
プライベート問合せにおける問合せ頻度を用いた制約緩和手法
 
プライバシを考慮した移動系列情報解析のための安全性の提案
プライバシを考慮した移動系列情報解析のための安全性の提案プライバシを考慮した移動系列情報解析のための安全性の提案
プライバシを考慮した移動系列情報解析のための安全性の提案
 
A Locality Sensitive Hashing Filter for Encrypted Vector Databases
A Locality Sensitive Hashing Filter for Encrypted Vector DatabasesA Locality Sensitive Hashing Filter for Encrypted Vector Databases
A Locality Sensitive Hashing Filter for Encrypted Vector Databases
 
位置情報解析のためのプライバシ保護手法
位置情報解析のためのプライバシ保護手法位置情報解析のためのプライバシ保護手法
位置情報解析のためのプライバシ保護手法
 
Sponsored Search Markets (from Networks, Crowds, and Markets: Reasoning About...
Sponsored Search Markets (from Networks, Crowds, and Markets: Reasoning About...Sponsored Search Markets (from Networks, Crowds, and Markets: Reasoning About...
Sponsored Search Markets (from Networks, Crowds, and Markets: Reasoning About...
 
Private Range Query by Perturbation and Matrix Based Encryption
Private Range Query by Perturbation and Matrix Based EncryptionPrivate Range Query by Perturbation and Matrix Based Encryption
Private Range Query by Perturbation and Matrix Based Encryption
 
暗号化データベースモデルにおける関係情報推定を防ぐ索引手法
暗号化データベースモデルにおける関係情報推定を防ぐ索引手法暗号化データベースモデルにおける関係情報推定を防ぐ索引手法
暗号化データベースモデルにおける関係情報推定を防ぐ索引手法
 
VLDB09勉強会 Session27 Privacy2
VLDB09勉強会 Session27 Privacy2VLDB09勉強会 Session27 Privacy2
VLDB09勉強会 Session27 Privacy2
 
Reducing Data Decryption Cost by Broadcast Encryption and Account Assignment ...
Reducing Data Decryption Cost by Broadcast Encryption and Account Assignment ...Reducing Data Decryption Cost by Broadcast Encryption and Account Assignment ...
Reducing Data Decryption Cost by Broadcast Encryption and Account Assignment ...
 
Security of Social Information from Query Analysis in DaaS
Security of Social Information from Query Analysis in DaaSSecurity of Social Information from Query Analysis in DaaS
Security of Social Information from Query Analysis in DaaS
 

Recently uploaded

SQL Database Design For Developers at php[tek] 2024
SQL Database Design For Developers at php[tek] 2024SQL Database Design For Developers at php[tek] 2024
SQL Database Design For Developers at php[tek] 2024Scott Keck-Warren
 
FULL ENJOY 🔝 8264348440 🔝 Call Girls in Diplomatic Enclave | Delhi
FULL ENJOY 🔝 8264348440 🔝 Call Girls in Diplomatic Enclave | DelhiFULL ENJOY 🔝 8264348440 🔝 Call Girls in Diplomatic Enclave | Delhi
FULL ENJOY 🔝 8264348440 🔝 Call Girls in Diplomatic Enclave | Delhisoniya singh
 
Breaking the Kubernetes Kill Chain: Host Path Mount
Breaking the Kubernetes Kill Chain: Host Path MountBreaking the Kubernetes Kill Chain: Host Path Mount
Breaking the Kubernetes Kill Chain: Host Path MountPuma Security, LLC
 
Install Stable Diffusion in windows machine
Install Stable Diffusion in windows machineInstall Stable Diffusion in windows machine
Install Stable Diffusion in windows machinePadma Pradeep
 
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 3652toLead Limited
 
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptxHampshireHUG
 
08448380779 Call Girls In Diplomatic Enclave Women Seeking Men
08448380779 Call Girls In Diplomatic Enclave Women Seeking Men08448380779 Call Girls In Diplomatic Enclave Women Seeking Men
08448380779 Call Girls In Diplomatic Enclave Women Seeking MenDelhi Call girls
 
AI as an Interface for Commercial Buildings
AI as an Interface for Commercial BuildingsAI as an Interface for Commercial Buildings
AI as an Interface for Commercial BuildingsMemoori
 
How to convert PDF to text with Nanonets
How to convert PDF to text with NanonetsHow to convert PDF to text with Nanonets
How to convert PDF to text with Nanonetsnaman860154
 
Slack Application Development 101 Slides
Slack Application Development 101 SlidesSlack Application Development 101 Slides
Slack Application Development 101 Slidespraypatel2
 
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...Neo4j
 
A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)Gabriella Davis
 
Key Features Of Token Development (1).pptx
Key  Features Of Token  Development (1).pptxKey  Features Of Token  Development (1).pptx
Key Features Of Token Development (1).pptxLBM Solutions
 
Scaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organizationScaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organizationRadu Cotescu
 
Factors to Consider When Choosing Accounts Payable Services Providers.pptx
Factors to Consider When Choosing Accounts Payable Services Providers.pptxFactors to Consider When Choosing Accounts Payable Services Providers.pptx
Factors to Consider When Choosing Accounts Payable Services Providers.pptxKatpro Technologies
 
IAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI SolutionsIAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI SolutionsEnterprise Knowledge
 
Pigging Solutions Piggable Sweeping Elbows
Pigging Solutions Piggable Sweeping ElbowsPigging Solutions Piggable Sweeping Elbows
Pigging Solutions Piggable Sweeping ElbowsPigging Solutions
 
Benefits Of Flutter Compared To Other Frameworks
Benefits Of Flutter Compared To Other FrameworksBenefits Of Flutter Compared To Other Frameworks
Benefits Of Flutter Compared To Other FrameworksSoftradix Technologies
 
WhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure service
WhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure serviceWhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure service
WhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure servicePooja Nehwal
 
Injustice - Developers Among Us (SciFiDevCon 2024)
Injustice - Developers Among Us (SciFiDevCon 2024)Injustice - Developers Among Us (SciFiDevCon 2024)
Injustice - Developers Among Us (SciFiDevCon 2024)Allon Mureinik
 

Recently uploaded (20)

SQL Database Design For Developers at php[tek] 2024
SQL Database Design For Developers at php[tek] 2024SQL Database Design For Developers at php[tek] 2024
SQL Database Design For Developers at php[tek] 2024
 
FULL ENJOY 🔝 8264348440 🔝 Call Girls in Diplomatic Enclave | Delhi
FULL ENJOY 🔝 8264348440 🔝 Call Girls in Diplomatic Enclave | DelhiFULL ENJOY 🔝 8264348440 🔝 Call Girls in Diplomatic Enclave | Delhi
FULL ENJOY 🔝 8264348440 🔝 Call Girls in Diplomatic Enclave | Delhi
 
Breaking the Kubernetes Kill Chain: Host Path Mount
Breaking the Kubernetes Kill Chain: Host Path MountBreaking the Kubernetes Kill Chain: Host Path Mount
Breaking the Kubernetes Kill Chain: Host Path Mount
 
Install Stable Diffusion in windows machine
Install Stable Diffusion in windows machineInstall Stable Diffusion in windows machine
Install Stable Diffusion in windows machine
 
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
 
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
 
08448380779 Call Girls In Diplomatic Enclave Women Seeking Men
08448380779 Call Girls In Diplomatic Enclave Women Seeking Men08448380779 Call Girls In Diplomatic Enclave Women Seeking Men
08448380779 Call Girls In Diplomatic Enclave Women Seeking Men
 
AI as an Interface for Commercial Buildings
AI as an Interface for Commercial BuildingsAI as an Interface for Commercial Buildings
AI as an Interface for Commercial Buildings
 
How to convert PDF to text with Nanonets
How to convert PDF to text with NanonetsHow to convert PDF to text with Nanonets
How to convert PDF to text with Nanonets
 
Slack Application Development 101 Slides
Slack Application Development 101 SlidesSlack Application Development 101 Slides
Slack Application Development 101 Slides
 
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...
 
A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)
 
Key Features Of Token Development (1).pptx
Key  Features Of Token  Development (1).pptxKey  Features Of Token  Development (1).pptx
Key Features Of Token Development (1).pptx
 
Scaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organizationScaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organization
 
Factors to Consider When Choosing Accounts Payable Services Providers.pptx
Factors to Consider When Choosing Accounts Payable Services Providers.pptxFactors to Consider When Choosing Accounts Payable Services Providers.pptx
Factors to Consider When Choosing Accounts Payable Services Providers.pptx
 
IAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI SolutionsIAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI Solutions
 
Pigging Solutions Piggable Sweeping Elbows
Pigging Solutions Piggable Sweeping ElbowsPigging Solutions Piggable Sweeping Elbows
Pigging Solutions Piggable Sweeping Elbows
 
Benefits Of Flutter Compared To Other Frameworks
Benefits Of Flutter Compared To Other FrameworksBenefits Of Flutter Compared To Other Frameworks
Benefits Of Flutter Compared To Other Frameworks
 
WhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure service
WhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure serviceWhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure service
WhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure service
 
Injustice - Developers Among Us (SciFiDevCon 2024)
Injustice - Developers Among Us (SciFiDevCon 2024)Injustice - Developers Among Us (SciFiDevCon 2024)
Injustice - Developers Among Us (SciFiDevCon 2024)
 

Frequency-based Constraint Relaxation for Private Cloud Database Queries

  • 1. Frequency-based Constraint Relaxation for Private Query Processing in Cloud Databases Junpei Kawamoto (Kyushu University, Japan) Patricia L. Gillett (École Polytechnique de Montréal)
  • 2. Cloud services and Privacy • Cloud services as cloud databases. • Sometimes people want to keep what they request to databases secret. May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 2 Find restaurants near by current location x. Location based services Want to read article x Forum sites
  • 3. Private query processing • Methodologies to obtain data w/o exposing queries. • Several protocols such as cPIR† & bbPIR†† are introduced. May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 3 Cloud Database Current location x query res. Encode x Compute query results w/o decoding queries †Kushilevitz, E. and Ostrovsky, R.: Replication Is Not Needed: Single Database, Computationally- Private Information Retrieval, Proc. of the 38th Annual Symposium on Foundations of Computer Science, pp. 364-373, 1997. E.g. Location Based Services ††Wang, S., Agrawal, D., and Abbadi, A.: Generalizing PIR for Practical Private Retrieval of Public Data, Proc. of the 24th Annual IFIP WG 11.3 Working Conference on Data and Applications Security and Privacy, pp. 1-16, 2010.
  • 4. Three ideas of private queries • We introduce three ideas for our discussion; • Search intention: what users hope to obtain from cloud services, • Query: request users send to servers to obtain data, • Handled set: data set which servers must check to compute the results. May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 4 Location x Want information associated with location x (search intention) query res. Cloud DB Must check these items to compute the result(Handled set) Handled set for x
  • 5. Existing private querying protocols • Most existing protocols impose two constraints: 1. Queries are encoded in such a way that servers can handle query processes but cannot actually decode queries; 2. Servers are made to check all data in the databases when computing any query result. • (2) means servers cannot distinguish any data in the DB. • Servers spend O(n) computational cost executing each query, where the database has n entries. May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 5 Cloud DB check all data any time query
  • 6. • There are cases in which we do not need to retrieve the entire database to sufficiently obscure search intentions. • It may be enough to hide where I am in downtown. • What area is enough to ensure our privacy? • E.g. • Servers may guess the search intention is x with high probability. • We should consider frequency of search intentions. Constraint Relaxation May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 6 shop x shop a shop b shop c Popular place Unpopular places Specified by a handled set a handled set
  • 7. Database model & frequency • Database model • Database D consists of n items: D = {t1, t2, …, tn}. • Users request x-th item tx (The search intention is x). • Handled set H(x) • The item set servers must check to compute the results of a given query associated with search intention x. • Frequency of search intentions • Freq(x) denotes the frequency of search intention x. • We assume Freq(x) is normalized. May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 7 Want tx (search intention) query res. DB H(x) Must check these items to compute the result(Handled set)
  • 8. Definition: Query Risk • A measure of exposure risk for private queries. • The query risk of search intention x and handled set H(x) is • Conditional probability that the search intention is x given that we know the handled set H used. • The more frequent x is, the higher the risk becomes. May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 8   )( )(Freq )(Freq ))(|(Risk xHy y x xHx frequenc y E.g. • Risk(3|{1,2,3,4,5}) = 3/9 • Risk(1|{1,2,3,4,5}) = 1/9 • Risk(1|{1,2}) = 1/3
  • 9. Definition: Privateness • The maximum risk of complete protocols: • We assume query risks should be less than or equals to the maximum risk of complete protocols: • Complete protocols are widely accepted that is the max. risk is, too. • Our approach will also be considered private if it satisfies this condition. May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 9 ))(|(Riskmax))(|(Risk completerelaxed yHyxHx Dy  ))(|(Riskmax complete xHx Dx i.e. w/o relaxation Our handled set Handled set for existing protocols
  • 10. Definition: Query processing cost • Query processing costs on servers. • We evaluate them by the size of handled sets. • The cost of search intention x is . May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 10 |)(|)(Cost xHx  query res. DB H(x) H(x) query res. DB vs.
  • 11. The problem • Find handled set H such that, for search intention x: 1. , 2. , 3. minimize Cost(x), 4. if multiple solutions have equal cost, chose the one maximizes • We solve this problem by Dynamic Programming. • O(n) algorithm but details are in our paper. • We also extend these problem and algorithm to • range queries in 1D data, • exact match queries in 2D data. May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 11 x Î H(x) )|(Riskmax))(|(Risk DyxHx Dy   )( )(FreqxHy y
  • 12. The protocol • Our protocol employs some existing protocol (PIR, etc.). • We assume the frequencies are public information. • User: whose search intension is x, 1. Compute optimized handled set H(x) using query frequencies. 2. Compute a private query for x assuming a DB has only items in H(x). 3. Send the query and H(x) to the cloud server. • The cloud server receiving the query, 1. Consider a sub-database consists of items in H(x). 2. Process the received query and return the result to the user. May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 12 query res. H(x) : sub-DB Cost: O(|H(x)|)
  • 13. • Dataset • Query logs from †. • Sampled 100,000 songs and 1,800,145 queries. Evaluation May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 13 †http://www.dtic.upf.edu/~ocelma/MusicRecommendationDataset/lastfm-1K.html Frequency of search intention x. x-axis: search intentions x y-axis: # of times users requested item tx
  • 14. Evaluation • Comparison of query risks. • In most cases, risks are bigger than those of the complete protocols. • Do not exceed the maximum risks of the complete protocol. May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 14 relaxed (avg.) is computed by   Dx xHxx ))(|(Risk)(Freq min. max. avg. comp. 5.6×10−7 3.9×10−3 1.0×10−5 relaxed 3.4×10−3 3.9×10−3 3.5×10−3 Query risk of search intention x.
  • 15. Evaluation • Comparison of query costs. • Our relaxation methodologies reduce costs in most cases. • The average cost is 6.5% that of complete protocols. May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 15 relaxed (avg.) is computed by Freq(x)´Cost(x) xÎD å min. max. avg. comp. 100000 100000 100000 relaxed 2 100000 6417 Query cost of search intention x.
  • 16. Conclusion • We introduced a frequency-based constraint relaxation methodology for private queries. • We relaxed constraint (2) of the complete protocols so that only a subset of the database is retrieved for each query. • We evaluated our proposal using a real dataset from . • Our protocol can reduce computational costs in servers in most cases, • The risk of a query being exposed is not bigger than the maximum risk in complete protocols. May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 16 Servers must check all data in the databases when computing any query result.
  • 17. Acknowledgement • This work is partly supported by • The Nakajima Foundation, • Artificial Intelligence Research Promotion Foundation, • Grant-in-Aid for Young Scientists (B) (26730065), Japan Society for the Promotion of Science (JSPS). May 6, 2014 Frequency-based Constraint Relaxation for Private Query Processing 17 Thank you for your attention!