SlideShare a Scribd company logo
Detection of Profile Injection Attacks
in
Recommender System
Presented By:
Ashish Pannu
Agenda
• Recommender System
- Introduction
- Types of Recommender System
- Why using Recommender System
• Profile Injection Attacks
- Why Attacker Attacks on Recommender System (Example)
- Attack Profile Structure
- Types of Attacks
o Push Attacks
o Nuke Attacks
- Attack Detection Attributes
o Generic Attributes
o Model Specific Attributes
• Proposed Approach
• Conclusion
• References
2
Recommender System
Introduction:
– Recommender System (RS) is based on information filter to predict
the rating or preference that a user would give to an unseen item [1].
– It is backbone of ecommernce websites (amazon, flipkart, myntra
etc.), social networking websites (facebook, linkedin, twitter, google +
etc.), matrimonial websites (shaadi.com, bharatmatrimonial.com etc.)
and many more.
3
Types of RS-
1. Collaborative Filtering (CF) based RS: based on correlation between
different users. It states that if two user had similar tastes in past will also
have same taste in future also.
– Most popular and widely used.
2. Content based RS: based on the information on the features(keywords)
of items rather than on the opinion of other users. E.g. in movie
recommendation keywords may be: movie name, actor, actress, genre
etc.
3. Knowledge based RS: customer defines his requirement explicitly. E.g. “I
want black color BMW car”.
Recommender System
4
Recommender System – why?
1. Value for the customer
– Narrow down the set of choice.
– Discover new things.
– Just exploring new items.
– Make shopping easier.
2. Value for the provider
– Increase user satisfaction.
– Increase the sell.
– Unique and personalized service to each customer.
– Obtain more knowledge about the customer.
– Opportunities for promotions.
5
Profile Injection Attack
Profile injection attack: Example
• Assume that user-user based CF is used.
• Pearson correlation as similarity measure.
• Neighborhood size of 1.
Item1 Item2 Item3 Item4 … Target Pearson
Ram 5 3 4 1 … ?
User1 3 1 2 5 … 5 -0.54
User2 4 3 3 3 … 2 0.68
User3 3 3 1 5 … 4 -0.72
User4 1 5 5 2 … 1 -0.02
6
Profile Injection Attack
Profile injection attack: Example
• Assume that user-user based CF is used.
• Pearson correlation as similarity measure.
• Neighborhood size of 1.
Item1 Item2 Item3 Item4 … Target Pearson
Ram 5 3 4 1 … ?
User1 3 1 2 5 … 5 -0.54
User2 4 3 3 3 … 2 0.68
User3 3 3 1 5 … 4 -0.72
User4 1 5 5 2 … 1 -0.02
User2 most similar to Ram
7
Profile Injection Attack
Profile injection attack: Example
• Assume that user-user based CF is used.
• Pearson correlation as similarity measure.
• Neighborhood size of 1.
User2 most similar to Ram
Item1 Item2 Item3 Item4 … Target Pearson
Ram 5 3 4 1 … ?
User1 3 1 2 5 … 5 -0.54
User2 4 3 3 3 … 2 0.68
User3 3 3 1 5 … 4 -0.72
User4 1 5 5 2 … 1 -0.02
Attack 5 3 4 3 … 5 0.87
Attack
8
Profile Injection Attack
Profile injection attack: Example
• Assume that user-user based CF is used.
• Pearson correlation as similarity measure.
• Neighborhood size of 1.
Item1 Item2 Item3 Item4 … Target Pearson
Ram 5 3 4 1 … ?
User1 3 1 2 5 … 5 -0.54
User2 4 3 3 3 … 2 0.68
User3 3 3 1 5 … 4 -0.72
User4 1 5 5 2 … 1 -0.02
Attack 5 3 4 3 … 5 0.87
Attack
Attack most similar to Ram
9
Attack Profile Structure
In order to look similar to genuine user, attacker gives ratings in a specific
manner so that it become hard to identify the attack profile [2].
Selected items: these items are chosen because of their association with the
target item.
Filler items: these items are randomly chosen and rating is given based upon
properties of attack.
Unrated items: No ratings are given to these items.
Target item: singleton item to which attacker promote or demote.
Item1 … ItemK … ItemN … ItemR Target
r_1 … r_k … r_l … r_n X
Selected items Filler items Unrated items Target item
10
Types of Attacks
Push Attacks- to promote a specific item. Maximum rating will be given to
the target item [3].
1. Random Attack:
– No rating to selected items.
– Average rating of system to the filler items.
– Least cost attack.
– Limited effect as compared to more advance attacks.
2. Average Attack:
– No rating to selected items.
– Average rating of item to the item of filler items.
– Additional cost of finding the average rating is involved.
– More effective as compared to random attack.
cont …
11
Types of Attacks
3. Bandwagon Attack:
– Maximum rating is given to popular items (selected items).
– Average rating of system to the filler items.
– It is a low cost attack.
4. Segment Attack:
– Maximum rating is given to the items of the same segment of target
item (selected items).
– Minimum rating is given to filler items.
– Best attack (impact wise) as compared to other attacks of same
category.
cont …
12
Types of Attacks
Nuke Attack- to demotion of an item. Minimum rating is given to the target
item.
1. Love/Hate Attack:
– No rating to selected items.
– Maximum rating to set of filler items.
2. Reverse Bandwagon Attack:
– Minimum rating is given to least popular items (selected item).
– Average rating of system is given to the set of filler items.
– More effective as compared to the love/hate attack.
We can not stop attacks, we can just increase the cost of attacks.
13
Attack Detection Attributes
Generic Attributes- based on general abnormal behavior of user [4]. It is
common for all attack types.
1. Rating Deviation from Mean Agreement (RDMA): finds profile’s average
rating deviation per item.
is the number of ratings given by user u. is rating given by user u
to item i. is average rating of item i. is number of ratings provided
for item i.
2. Degree Similarity with Top Neighbors (DegSim): find the average
similarity of profile with top k neighbors.
Nu r i,u
ri ti
14
Attack Detection Attributes
3. Length Variance: finds the variance in the length (# ratings given) of a given
profile from the average length of database.
Model Specific Attributes- focuses on the signature of attacks type.
1. Mean Variance (MeanVar): used for average and random attack.
is the set of items rated by user u.
2. Filler Mean Target Difference (FMTD): used for bandwagon, reverse
bandwagon and segment attack.
is the set of filler items rated by user u.
pu
p f,u
15
Evaluation Metrics
1. Precision: count total number of profiles that are labeled as attack [4].
2. Recall: count total number of actual attack in the system.
Reality
Prediction
Actually Attack Actually Good
Rated Attack True Positive (Tp) False Positive (Fp)
Rated Good False Negative (Fn) True Negative (Tn)
16
Proposed Approach
Steps-
1. Calculate the attack detection attributes (generic and model specific
attributes).
2. Apply statistical models using k-fold cross validation
3. Compare the accuracy of statistical models.
4. Pick the top three performing models.
5. Ensemble the models (resulting models from step 4) using voting
approach.
Data Set Used: MovieLens-100K
– 943 users
– 1682 movies
– 100000 ratings
17
Performance Analysis
Table 1: Performance analysis of 10% average attack.
Table 2: Performance analysis for bandwagon attack at 5% filler size.
Filler Size 1% 10% 20% 30% 40% 50%
Models P R P R P R P R P R P R
Decision Tree .90 .892 .921 .93 .928 .919 .939 .912 .94 .94 .961 .968
Random Forest .929 .930 .939 .92 .948 .948 .952 .956 .962 .961 .973 .979
Ada Boost .9 .908 .914 .918 .93 .924 .935 .938 .948 .934 .950 .943
SVM .938 .927 .949 .94 .959 .959 .971 .975 .979 .981 .988 .989
Linear Regression .89 .862 .89 .907 .918 .91 .925 .918 .929 .902 .93 .931
Neural Network .949 .938 .95 .943 .951 .954 .959 .958 .968 .967 .979 .98
Ensemble .939 .932 .946 .934 .953 .953 .961 .963 .968 .969 .98 .982
Attack Size 1% 3% 6% 9% 12% 15%
Models P R P R P R P R P R P R
Decision Tree .822 .811 .844 .830 .859 .848 .868 .855 .889 .872 .901 .928
Random Forest .880 .872 .894 .897 .905 .912 .909 .902 .929 .917 .957 .922
Ada Boost .819 .802 .829 .810 .834 .828 .858 .840 .883 .868 .902 .915
SVM .901 .902 .912 .928 .928 .908 .937 .911 .972 .959 .99 .984
Linear Regression .862 .842 .872 .882 .882 .908 .908 .919 .918 .935 .939 .941
Neural Network .892 .901 .919 .911 .922 .929 .959 .958 .961 .969 .97 .964
Ensemble .891 .892 .908 .912 .918 .916 .935 .924 .954 .948 .972 .957
18
Conclusion
1. I tried to explain why RS is so important now a days for ecommerce.
2. I also tried to explain, how a attacker can manipulate the results of RS.
3. I focused on several keys areas of attacks against recommender system
i.e. different types of attacks, attack detection attributes and model
evaluation metrics.
4. I found that random forest, SVM and neural networks perform better
than other models in attack detection.
5. I present ensemble approach for the attack detection and I found that
performance is not best in any case but it also does not give worst
performance in any case.
19
20
References
[1]. Davoodi, Fatemeh Ghiyafeh, and Omid Fatemi. "Tag based recommender
system for social bookmarking sites." In Advances in Social Networks
Analysis and Mining (ASONAM), 2012 IEEE/ACM International
Conference on, pp. 934-940. IEEE, 2012.
[2]. Lam, Shyong K., and John Riedl. "Shilling recommender systems for fun
and profit." In Proceedings of the 13th international conference on
World Wide Web, pp. 393-402. ACM, 2004.
[3]. Mobasher, Bamshad, Robin Burke, Runa Bhaumik, and Chad Williams.
"Toward trustworthy recommender systems: An analysis of attack
models and algorithm robustness." ACM Transactions on Internet
Technology (TOIT) 7, no. 4 (2007): 23.
[4]. O'Mahony, Michael, Neil Hurley, Nicholas Kushmerick, and Guénolé
Silvestre. "Collaborative recommendation: A robustness analysis." ACM
Transactions on Internet Technology (TOIT) 4, no. 4 (2004): 344-377.
21

More Related Content

What's hot

Adversarial Attacks for Recommender Systems
Adversarial Attacks for Recommender SystemsAdversarial Attacks for Recommender Systems
Adversarial Attacks for Recommender Systems
WQ Fan
 
Movie recommendation system using collaborative filtering system
Movie recommendation system using collaborative filtering system Movie recommendation system using collaborative filtering system
Movie recommendation system using collaborative filtering system
Mauryasuraj98
 
Object detection presentation
Object detection presentationObject detection presentation
Object detection presentation
AshwinBicholiya
 
Image segmentation with deep learning
Image segmentation with deep learningImage segmentation with deep learning
Image segmentation with deep learning
Antonio Rueda-Toicen
 
Recommender system
Recommender systemRecommender system
Recommender system
Saiguru P.v
 
Recommender Systems
Recommender SystemsRecommender Systems
Recommender Systems
Girish Khanzode
 
Recommender systems using collaborative filtering
Recommender systems using collaborative filteringRecommender systems using collaborative filtering
Recommender systems using collaborative filtering
D Yogendra Rao
 
Computer Vision - Real Time Face Recognition using Open CV and Python
Computer Vision - Real Time Face Recognition using Open CV and PythonComputer Vision - Real Time Face Recognition using Open CV and Python
Computer Vision - Real Time Face Recognition using Open CV and Python
Akash Satamkar
 
Face recognition technology
Face recognition technologyFace recognition technology
Face recognition technology
ranjit banshpal
 
Movies Recommendation System
Movies Recommendation SystemMovies Recommendation System
Movies Recommendation System
Shubham Patil
 
Movie Recommendation engine
Movie Recommendation engineMovie Recommendation engine
Movie Recommendation engine
Jayesh Lahori
 
Introduction to Deep face detection and recognition
Introduction to Deep face detection and recognitionIntroduction to Deep face detection and recognition
Introduction to Deep face detection and recognition
Apache MXNet
 
Collaborative Filtering using KNN
Collaborative Filtering using KNNCollaborative Filtering using KNN
Collaborative Filtering using KNN
Şeyda Hatipoğlu
 
Movies recommendation system in R Studio, Machine learning
Movies recommendation system in  R Studio, Machine learning Movies recommendation system in  R Studio, Machine learning
Movies recommendation system in R Studio, Machine learning
Mauryasuraj98
 
Driver drowsiness monitoring system using visual behavior and Machine Learning.
Driver drowsiness monitoring system using visual behavior and Machine Learning.Driver drowsiness monitoring system using visual behavior and Machine Learning.
Driver drowsiness monitoring system using visual behavior and Machine Learning.
AasimAhmedKhanJawaad
 
Content based filtering
Content based filteringContent based filtering
Content based filtering
Bendito Freitas Ribeiro
 
Object Detection & Tracking
Object Detection & TrackingObject Detection & Tracking
Object Detection & Tracking
Akshay Gujarathi
 
cnn ppt.pptx
cnn ppt.pptxcnn ppt.pptx
cnn ppt.pptx
rohithprabhas1
 
Nearest Neighbor Algorithm Zaffar Ahmed
Nearest Neighbor Algorithm  Zaffar AhmedNearest Neighbor Algorithm  Zaffar Ahmed
Nearest Neighbor Algorithm Zaffar AhmedZaffar Ahmed Shaikh
 
K - Nearest neighbor ( KNN )
K - Nearest neighbor  ( KNN )K - Nearest neighbor  ( KNN )
K - Nearest neighbor ( KNN )
Mohammad Junaid Khan
 

What's hot (20)

Adversarial Attacks for Recommender Systems
Adversarial Attacks for Recommender SystemsAdversarial Attacks for Recommender Systems
Adversarial Attacks for Recommender Systems
 
Movie recommendation system using collaborative filtering system
Movie recommendation system using collaborative filtering system Movie recommendation system using collaborative filtering system
Movie recommendation system using collaborative filtering system
 
Object detection presentation
Object detection presentationObject detection presentation
Object detection presentation
 
Image segmentation with deep learning
Image segmentation with deep learningImage segmentation with deep learning
Image segmentation with deep learning
 
Recommender system
Recommender systemRecommender system
Recommender system
 
Recommender Systems
Recommender SystemsRecommender Systems
Recommender Systems
 
Recommender systems using collaborative filtering
Recommender systems using collaborative filteringRecommender systems using collaborative filtering
Recommender systems using collaborative filtering
 
Computer Vision - Real Time Face Recognition using Open CV and Python
Computer Vision - Real Time Face Recognition using Open CV and PythonComputer Vision - Real Time Face Recognition using Open CV and Python
Computer Vision - Real Time Face Recognition using Open CV and Python
 
Face recognition technology
Face recognition technologyFace recognition technology
Face recognition technology
 
Movies Recommendation System
Movies Recommendation SystemMovies Recommendation System
Movies Recommendation System
 
Movie Recommendation engine
Movie Recommendation engineMovie Recommendation engine
Movie Recommendation engine
 
Introduction to Deep face detection and recognition
Introduction to Deep face detection and recognitionIntroduction to Deep face detection and recognition
Introduction to Deep face detection and recognition
 
Collaborative Filtering using KNN
Collaborative Filtering using KNNCollaborative Filtering using KNN
Collaborative Filtering using KNN
 
Movies recommendation system in R Studio, Machine learning
Movies recommendation system in  R Studio, Machine learning Movies recommendation system in  R Studio, Machine learning
Movies recommendation system in R Studio, Machine learning
 
Driver drowsiness monitoring system using visual behavior and Machine Learning.
Driver drowsiness monitoring system using visual behavior and Machine Learning.Driver drowsiness monitoring system using visual behavior and Machine Learning.
Driver drowsiness monitoring system using visual behavior and Machine Learning.
 
Content based filtering
Content based filteringContent based filtering
Content based filtering
 
Object Detection & Tracking
Object Detection & TrackingObject Detection & Tracking
Object Detection & Tracking
 
cnn ppt.pptx
cnn ppt.pptxcnn ppt.pptx
cnn ppt.pptx
 
Nearest Neighbor Algorithm Zaffar Ahmed
Nearest Neighbor Algorithm  Zaffar AhmedNearest Neighbor Algorithm  Zaffar Ahmed
Nearest Neighbor Algorithm Zaffar Ahmed
 
K - Nearest neighbor ( KNN )
K - Nearest neighbor  ( KNN )K - Nearest neighbor  ( KNN )
K - Nearest neighbor ( KNN )
 

Viewers also liked

Recommender Systems in E-Commerce
Recommender Systems in E-CommerceRecommender Systems in E-Commerce
Recommender Systems in E-CommerceRoger Chen
 
Recommender system algorithm and architecture
Recommender system algorithm and architectureRecommender system algorithm and architecture
Recommender system algorithm and architectureLiang Xiang
 
Profile Injection Attack Detection in Recommender System
Profile Injection Attack Detection in Recommender SystemProfile Injection Attack Detection in Recommender System
Profile Injection Attack Detection in Recommender SystemASHISH PANNU
 
Spatially Aware Recommendation System
Spatially Aware Recommendation SystemSpatially Aware Recommendation System
Spatially Aware Recommendation SystemVeer Chandra
 
[Decisions2013@RecSys]The Role of Emotions in Context-aware Recommendation
[Decisions2013@RecSys]The Role of Emotions in Context-aware Recommendation[Decisions2013@RecSys]The Role of Emotions in Context-aware Recommendation
[Decisions2013@RecSys]The Role of Emotions in Context-aware Recommendation
YONG ZHENG
 
Comparison of Techniques for Measuring Research Coverage of Scientific Papers...
Comparison of Techniques for Measuring Research Coverage of Scientific Papers...Comparison of Techniques for Measuring Research Coverage of Scientific Papers...
Comparison of Techniques for Measuring Research Coverage of Scientific Papers...
Aravind Sesagiri Raamkumar
 
Solving the AL Chicken-and-Egg Corpus and Model Problem
Solving the AL Chicken-and-Egg Corpus and Model ProblemSolving the AL Chicken-and-Egg Corpus and Model Problem
Solving the AL Chicken-and-Egg Corpus and Model Problem
Dain Kaplan
 
Recommender.system.presentation.pjug.01.21.2014
Recommender.system.presentation.pjug.01.21.2014Recommender.system.presentation.pjug.01.21.2014
Recommender.system.presentation.pjug.01.21.2014
rpbrehm
 
II-SDV 2014 Recommender Systems for Analysis Applications (Roger Bradford - A...
II-SDV 2014 Recommender Systems for Analysis Applications (Roger Bradford - A...II-SDV 2014 Recommender Systems for Analysis Applications (Roger Bradford - A...
II-SDV 2014 Recommender Systems for Analysis Applications (Roger Bradford - A...Dr. Haxel Consult
 
Recommendation system
Recommendation systemRecommendation system
Recommendation system
Rishabh Mehta
 
Toward the Next Generation of Recommender Systems:
Toward the Next Generation of Recommender Systems: Toward the Next Generation of Recommender Systems:
Toward the Next Generation of Recommender Systems: Vincent Chu
 
Multi Criteria Recommender Systems - Overview
Multi Criteria Recommender Systems - OverviewMulti Criteria Recommender Systems - Overview
Multi Criteria Recommender Systems - Overview
Davide Giannico
 
Offline evaluation of recommender systems: all pain and no gain?
Offline evaluation of recommender systems: all pain and no gain?Offline evaluation of recommender systems: all pain and no gain?
Offline evaluation of recommender systems: all pain and no gain?
Mark Levy
 
Your own recommendation engine with neo4j and reco4php - DPC16
Your own recommendation engine with neo4j and reco4php - DPC16Your own recommendation engine with neo4j and reco4php - DPC16
Your own recommendation engine with neo4j and reco4php - DPC16
Christophe Willemsen
 
Summary of a Recommender Systems Survey paper
Summary of a Recommender Systems Survey paperSummary of a Recommender Systems Survey paper
Summary of a Recommender Systems Survey paper
Changsung Moon
 
Recommendation Engine Project Presentation
Recommendation Engine Project PresentationRecommendation Engine Project Presentation
Recommendation Engine Project Presentation
19Divya
 
Impersonal Recommendation system on top of Hadoop
Impersonal Recommendation system on top of HadoopImpersonal Recommendation system on top of Hadoop
Impersonal Recommendation system on top of Hadoop
Kostiantyn Kudriavtsev
 
Recommender Systems and Active Learning
Recommender Systems and Active LearningRecommender Systems and Active Learning
Recommender Systems and Active Learning
Dain Kaplan
 
Online recommendations at scale using matrix factorisation
Online recommendations at scale using matrix factorisationOnline recommendations at scale using matrix factorisation
Online recommendations at scale using matrix factorisation
Marcus Ljungblad
 

Viewers also liked (20)

Recommender Systems in E-Commerce
Recommender Systems in E-CommerceRecommender Systems in E-Commerce
Recommender Systems in E-Commerce
 
Recommender system algorithm and architecture
Recommender system algorithm and architectureRecommender system algorithm and architecture
Recommender system algorithm and architecture
 
Profile Injection Attack Detection in Recommender System
Profile Injection Attack Detection in Recommender SystemProfile Injection Attack Detection in Recommender System
Profile Injection Attack Detection in Recommender System
 
Spatially Aware Recommendation System
Spatially Aware Recommendation SystemSpatially Aware Recommendation System
Spatially Aware Recommendation System
 
[Decisions2013@RecSys]The Role of Emotions in Context-aware Recommendation
[Decisions2013@RecSys]The Role of Emotions in Context-aware Recommendation[Decisions2013@RecSys]The Role of Emotions in Context-aware Recommendation
[Decisions2013@RecSys]The Role of Emotions in Context-aware Recommendation
 
Comparison of Techniques for Measuring Research Coverage of Scientific Papers...
Comparison of Techniques for Measuring Research Coverage of Scientific Papers...Comparison of Techniques for Measuring Research Coverage of Scientific Papers...
Comparison of Techniques for Measuring Research Coverage of Scientific Papers...
 
Solving the AL Chicken-and-Egg Corpus and Model Problem
Solving the AL Chicken-and-Egg Corpus and Model ProblemSolving the AL Chicken-and-Egg Corpus and Model Problem
Solving the AL Chicken-and-Egg Corpus and Model Problem
 
Recommender.system.presentation.pjug.01.21.2014
Recommender.system.presentation.pjug.01.21.2014Recommender.system.presentation.pjug.01.21.2014
Recommender.system.presentation.pjug.01.21.2014
 
II-SDV 2014 Recommender Systems for Analysis Applications (Roger Bradford - A...
II-SDV 2014 Recommender Systems for Analysis Applications (Roger Bradford - A...II-SDV 2014 Recommender Systems for Analysis Applications (Roger Bradford - A...
II-SDV 2014 Recommender Systems for Analysis Applications (Roger Bradford - A...
 
Recommendation system
Recommendation systemRecommendation system
Recommendation system
 
Toward the Next Generation of Recommender Systems:
Toward the Next Generation of Recommender Systems: Toward the Next Generation of Recommender Systems:
Toward the Next Generation of Recommender Systems:
 
Multi Criteria Recommender Systems - Overview
Multi Criteria Recommender Systems - OverviewMulti Criteria Recommender Systems - Overview
Multi Criteria Recommender Systems - Overview
 
Offline evaluation of recommender systems: all pain and no gain?
Offline evaluation of recommender systems: all pain and no gain?Offline evaluation of recommender systems: all pain and no gain?
Offline evaluation of recommender systems: all pain and no gain?
 
Your own recommendation engine with neo4j and reco4php - DPC16
Your own recommendation engine with neo4j and reco4php - DPC16Your own recommendation engine with neo4j and reco4php - DPC16
Your own recommendation engine with neo4j and reco4php - DPC16
 
Summary of a Recommender Systems Survey paper
Summary of a Recommender Systems Survey paperSummary of a Recommender Systems Survey paper
Summary of a Recommender Systems Survey paper
 
Recommender systems
Recommender systemsRecommender systems
Recommender systems
 
Recommendation Engine Project Presentation
Recommendation Engine Project PresentationRecommendation Engine Project Presentation
Recommendation Engine Project Presentation
 
Impersonal Recommendation system on top of Hadoop
Impersonal Recommendation system on top of HadoopImpersonal Recommendation system on top of Hadoop
Impersonal Recommendation system on top of Hadoop
 
Recommender Systems and Active Learning
Recommender Systems and Active LearningRecommender Systems and Active Learning
Recommender Systems and Active Learning
 
Online recommendations at scale using matrix factorisation
Online recommendations at scale using matrix factorisationOnline recommendations at scale using matrix factorisation
Online recommendations at scale using matrix factorisation
 

Similar to Profile injection attack detection in recommender system

Major
MajorMajor
Movie Recommender System Using Artificial Intelligence
Movie Recommender System Using Artificial Intelligence Movie Recommender System Using Artificial Intelligence
Movie Recommender System Using Artificial Intelligence
Shrutika Oswal
 
Developing a Secured Recommender System in Social Semantic Network
Developing a Secured Recommender System in Social Semantic NetworkDeveloping a Secured Recommender System in Social Semantic Network
Developing a Secured Recommender System in Social Semantic Network
Tamer Rezk
 
Strategies of detecting Profile-injection attacks in E-Commerce Recommender S...
Strategies of detecting Profile-injection attacks in E-Commerce Recommender S...Strategies of detecting Profile-injection attacks in E-Commerce Recommender S...
Strategies of detecting Profile-injection attacks in E-Commerce Recommender S...
IJERA Editor
 
An Example of use the Threat Modeling Tool (FFRI Monthly Research Nov 2016)
An Example of use the Threat Modeling Tool (FFRI Monthly Research Nov 2016)An Example of use the Threat Modeling Tool (FFRI Monthly Research Nov 2016)
An Example of use the Threat Modeling Tool (FFRI Monthly Research Nov 2016)
FFRI, Inc.
 
An Overview of Advesarial-attack-in-Recommender-system.pptx
An Overview of Advesarial-attack-in-Recommender-system.pptxAn Overview of Advesarial-attack-in-Recommender-system.pptx
An Overview of Advesarial-attack-in-Recommender-system.pptx
vudinhphuong96
 
Recommender Systems from A to Z – The Right Dataset
Recommender Systems from A to Z – The Right DatasetRecommender Systems from A to Z – The Right Dataset
Recommender Systems from A to Z – The Right Dataset
Crossing Minds
 
Recommendation Systems
Recommendation SystemsRecommendation Systems
Recommendation Systems
Robin Reni
 
Security evaluation of pattern classifiers under attack
Security evaluation of pattern classifiers under attackSecurity evaluation of pattern classifiers under attack
Security evaluation of pattern classifiers under attack
Shakas Technologies
 
Self-Learning Systems for Cyber Security
Self-Learning Systems for Cyber SecuritySelf-Learning Systems for Cyber Security
Self-Learning Systems for Cyber Security
Kim Hammar
 
JPJ1425 Security Evaluation of Pattern Classifiers under Attack
JPJ1425  Security Evaluation of Pattern Classifiers under AttackJPJ1425  Security Evaluation of Pattern Classifiers under Attack
JPJ1425 Security Evaluation of Pattern Classifiers under Attack
chennaijp
 
Aaa ped-21-Recommender Systems: Content-based Filtering
Aaa ped-21-Recommender Systems: Content-based FilteringAaa ped-21-Recommender Systems: Content-based Filtering
Aaa ped-21-Recommender Systems: Content-based Filtering
AminaRepo
 
Injection Attack detection using ML for
Injection Attack detection using ML  forInjection Attack detection using ML  for
Injection Attack detection using ML for
Khazane Hassan
 
Self-learning systems for cyber security
Self-learning systems for cyber securitySelf-learning systems for cyber security
Self-learning systems for cyber security
Kim Hammar
 
The Diamond Model for Intrusion Analysis - Threat Intelligence
The Diamond Model for Intrusion Analysis - Threat IntelligenceThe Diamond Model for Intrusion Analysis - Threat Intelligence
The Diamond Model for Intrusion Analysis - Threat Intelligence
ThreatConnect
 
The security mindset securing social media integrations and social learning...
The security mindset   securing social media integrations and social learning...The security mindset   securing social media integrations and social learning...
The security mindset securing social media integrations and social learning...franco_bb
 
Core Insight Enterprise 5min
Core Insight Enterprise 5minCore Insight Enterprise 5min
Core Insight Enterprise 5minNsolera
 
Collaborative Filtering Recommendation System
Collaborative Filtering Recommendation SystemCollaborative Filtering Recommendation System
Collaborative Filtering Recommendation System
Milind Gokhale
 
Network Based Intrusion Detection System using Filter Based Feature Selection...
Network Based Intrusion Detection System using Filter Based Feature Selection...Network Based Intrusion Detection System using Filter Based Feature Selection...
Network Based Intrusion Detection System using Filter Based Feature Selection...
IRJET Journal
 

Similar to Profile injection attack detection in recommender system (20)

presentation
presentationpresentation
presentation
 
Major
MajorMajor
Major
 
Movie Recommender System Using Artificial Intelligence
Movie Recommender System Using Artificial Intelligence Movie Recommender System Using Artificial Intelligence
Movie Recommender System Using Artificial Intelligence
 
Developing a Secured Recommender System in Social Semantic Network
Developing a Secured Recommender System in Social Semantic NetworkDeveloping a Secured Recommender System in Social Semantic Network
Developing a Secured Recommender System in Social Semantic Network
 
Strategies of detecting Profile-injection attacks in E-Commerce Recommender S...
Strategies of detecting Profile-injection attacks in E-Commerce Recommender S...Strategies of detecting Profile-injection attacks in E-Commerce Recommender S...
Strategies of detecting Profile-injection attacks in E-Commerce Recommender S...
 
An Example of use the Threat Modeling Tool (FFRI Monthly Research Nov 2016)
An Example of use the Threat Modeling Tool (FFRI Monthly Research Nov 2016)An Example of use the Threat Modeling Tool (FFRI Monthly Research Nov 2016)
An Example of use the Threat Modeling Tool (FFRI Monthly Research Nov 2016)
 
An Overview of Advesarial-attack-in-Recommender-system.pptx
An Overview of Advesarial-attack-in-Recommender-system.pptxAn Overview of Advesarial-attack-in-Recommender-system.pptx
An Overview of Advesarial-attack-in-Recommender-system.pptx
 
Recommender Systems from A to Z – The Right Dataset
Recommender Systems from A to Z – The Right DatasetRecommender Systems from A to Z – The Right Dataset
Recommender Systems from A to Z – The Right Dataset
 
Recommendation Systems
Recommendation SystemsRecommendation Systems
Recommendation Systems
 
Security evaluation of pattern classifiers under attack
Security evaluation of pattern classifiers under attackSecurity evaluation of pattern classifiers under attack
Security evaluation of pattern classifiers under attack
 
Self-Learning Systems for Cyber Security
Self-Learning Systems for Cyber SecuritySelf-Learning Systems for Cyber Security
Self-Learning Systems for Cyber Security
 
JPJ1425 Security Evaluation of Pattern Classifiers under Attack
JPJ1425  Security Evaluation of Pattern Classifiers under AttackJPJ1425  Security Evaluation of Pattern Classifiers under Attack
JPJ1425 Security Evaluation of Pattern Classifiers under Attack
 
Aaa ped-21-Recommender Systems: Content-based Filtering
Aaa ped-21-Recommender Systems: Content-based FilteringAaa ped-21-Recommender Systems: Content-based Filtering
Aaa ped-21-Recommender Systems: Content-based Filtering
 
Injection Attack detection using ML for
Injection Attack detection using ML  forInjection Attack detection using ML  for
Injection Attack detection using ML for
 
Self-learning systems for cyber security
Self-learning systems for cyber securitySelf-learning systems for cyber security
Self-learning systems for cyber security
 
The Diamond Model for Intrusion Analysis - Threat Intelligence
The Diamond Model for Intrusion Analysis - Threat IntelligenceThe Diamond Model for Intrusion Analysis - Threat Intelligence
The Diamond Model for Intrusion Analysis - Threat Intelligence
 
The security mindset securing social media integrations and social learning...
The security mindset   securing social media integrations and social learning...The security mindset   securing social media integrations and social learning...
The security mindset securing social media integrations and social learning...
 
Core Insight Enterprise 5min
Core Insight Enterprise 5minCore Insight Enterprise 5min
Core Insight Enterprise 5min
 
Collaborative Filtering Recommendation System
Collaborative Filtering Recommendation SystemCollaborative Filtering Recommendation System
Collaborative Filtering Recommendation System
 
Network Based Intrusion Detection System using Filter Based Feature Selection...
Network Based Intrusion Detection System using Filter Based Feature Selection...Network Based Intrusion Detection System using Filter Based Feature Selection...
Network Based Intrusion Detection System using Filter Based Feature Selection...
 

Recently uploaded

一比一原版(YU毕业证)约克大学毕业证成绩单
一比一原版(YU毕业证)约克大学毕业证成绩单一比一原版(YU毕业证)约克大学毕业证成绩单
一比一原版(YU毕业证)约克大学毕业证成绩单
enxupq
 
一比一原版(IIT毕业证)伊利诺伊理工大学毕业证成绩单
一比一原版(IIT毕业证)伊利诺伊理工大学毕业证成绩单一比一原版(IIT毕业证)伊利诺伊理工大学毕业证成绩单
一比一原版(IIT毕业证)伊利诺伊理工大学毕业证成绩单
ewymefz
 
Opendatabay - Open Data Marketplace.pptx
Opendatabay - Open Data Marketplace.pptxOpendatabay - Open Data Marketplace.pptx
Opendatabay - Open Data Marketplace.pptx
Opendatabay
 
SOCRadar Germany 2024 Threat Landscape Report
SOCRadar Germany 2024 Threat Landscape ReportSOCRadar Germany 2024 Threat Landscape Report
SOCRadar Germany 2024 Threat Landscape Report
SOCRadar
 
一比一原版(CBU毕业证)不列颠海角大学毕业证成绩单
一比一原版(CBU毕业证)不列颠海角大学毕业证成绩单一比一原版(CBU毕业证)不列颠海角大学毕业证成绩单
一比一原版(CBU毕业证)不列颠海角大学毕业证成绩单
nscud
 
Chatty Kathy - UNC Bootcamp Final Project Presentation - Final Version - 5.23...
Chatty Kathy - UNC Bootcamp Final Project Presentation - Final Version - 5.23...Chatty Kathy - UNC Bootcamp Final Project Presentation - Final Version - 5.23...
Chatty Kathy - UNC Bootcamp Final Project Presentation - Final Version - 5.23...
John Andrews
 
一比一原版(UVic毕业证)维多利亚大学毕业证成绩单
一比一原版(UVic毕业证)维多利亚大学毕业证成绩单一比一原版(UVic毕业证)维多利亚大学毕业证成绩单
一比一原版(UVic毕业证)维多利亚大学毕业证成绩单
ukgaet
 
原版制作(Deakin毕业证书)迪肯大学毕业证学位证一模一样
原版制作(Deakin毕业证书)迪肯大学毕业证学位证一模一样原版制作(Deakin毕业证书)迪肯大学毕业证学位证一模一样
原版制作(Deakin毕业证书)迪肯大学毕业证学位证一模一样
u86oixdj
 
社内勉強会資料_LLM Agents                              .
社内勉強会資料_LLM Agents                              .社内勉強会資料_LLM Agents                              .
社内勉強会資料_LLM Agents                              .
NABLAS株式会社
 
FP Growth Algorithm and its Applications
FP Growth Algorithm and its ApplicationsFP Growth Algorithm and its Applications
FP Growth Algorithm and its Applications
MaleehaSheikh2
 
一比一原版(CU毕业证)卡尔顿大学毕业证成绩单
一比一原版(CU毕业证)卡尔顿大学毕业证成绩单一比一原版(CU毕业证)卡尔顿大学毕业证成绩单
一比一原版(CU毕业证)卡尔顿大学毕业证成绩单
yhkoc
 
一比一原版(RUG毕业证)格罗宁根大学毕业证成绩单
一比一原版(RUG毕业证)格罗宁根大学毕业证成绩单一比一原版(RUG毕业证)格罗宁根大学毕业证成绩单
一比一原版(RUG毕业证)格罗宁根大学毕业证成绩单
vcaxypu
 
一比一原版(UofS毕业证书)萨省大学毕业证如何办理
一比一原版(UofS毕业证书)萨省大学毕业证如何办理一比一原版(UofS毕业证书)萨省大学毕业证如何办理
一比一原版(UofS毕业证书)萨省大学毕业证如何办理
v3tuleee
 
Empowering Data Analytics Ecosystem.pptx
Empowering Data Analytics Ecosystem.pptxEmpowering Data Analytics Ecosystem.pptx
Empowering Data Analytics Ecosystem.pptx
benishzehra469
 
一比一原版(Adelaide毕业证书)阿德莱德大学毕业证如何办理
一比一原版(Adelaide毕业证书)阿德莱德大学毕业证如何办理一比一原版(Adelaide毕业证书)阿德莱德大学毕业证如何办理
一比一原版(Adelaide毕业证书)阿德莱德大学毕业证如何办理
slg6lamcq
 
一比一原版(UPenn毕业证)宾夕法尼亚大学毕业证成绩单
一比一原版(UPenn毕业证)宾夕法尼亚大学毕业证成绩单一比一原版(UPenn毕业证)宾夕法尼亚大学毕业证成绩单
一比一原版(UPenn毕业证)宾夕法尼亚大学毕业证成绩单
ewymefz
 
一比一原版(CBU毕业证)卡普顿大学毕业证如何办理
一比一原版(CBU毕业证)卡普顿大学毕业证如何办理一比一原版(CBU毕业证)卡普顿大学毕业证如何办理
一比一原版(CBU毕业证)卡普顿大学毕业证如何办理
ahzuo
 
一比一原版(TWU毕业证)西三一大学毕业证成绩单
一比一原版(TWU毕业证)西三一大学毕业证成绩单一比一原版(TWU毕业证)西三一大学毕业证成绩单
一比一原版(TWU毕业证)西三一大学毕业证成绩单
ocavb
 
Algorithmic optimizations for Dynamic Levelwise PageRank (from STICD) : SHORT...
Algorithmic optimizations for Dynamic Levelwise PageRank (from STICD) : SHORT...Algorithmic optimizations for Dynamic Levelwise PageRank (from STICD) : SHORT...
Algorithmic optimizations for Dynamic Levelwise PageRank (from STICD) : SHORT...
Subhajit Sahu
 
Ch03-Managing the Object-Oriented Information Systems Project a.pdf
Ch03-Managing the Object-Oriented Information Systems Project a.pdfCh03-Managing the Object-Oriented Information Systems Project a.pdf
Ch03-Managing the Object-Oriented Information Systems Project a.pdf
haila53
 

Recently uploaded (20)

一比一原版(YU毕业证)约克大学毕业证成绩单
一比一原版(YU毕业证)约克大学毕业证成绩单一比一原版(YU毕业证)约克大学毕业证成绩单
一比一原版(YU毕业证)约克大学毕业证成绩单
 
一比一原版(IIT毕业证)伊利诺伊理工大学毕业证成绩单
一比一原版(IIT毕业证)伊利诺伊理工大学毕业证成绩单一比一原版(IIT毕业证)伊利诺伊理工大学毕业证成绩单
一比一原版(IIT毕业证)伊利诺伊理工大学毕业证成绩单
 
Opendatabay - Open Data Marketplace.pptx
Opendatabay - Open Data Marketplace.pptxOpendatabay - Open Data Marketplace.pptx
Opendatabay - Open Data Marketplace.pptx
 
SOCRadar Germany 2024 Threat Landscape Report
SOCRadar Germany 2024 Threat Landscape ReportSOCRadar Germany 2024 Threat Landscape Report
SOCRadar Germany 2024 Threat Landscape Report
 
一比一原版(CBU毕业证)不列颠海角大学毕业证成绩单
一比一原版(CBU毕业证)不列颠海角大学毕业证成绩单一比一原版(CBU毕业证)不列颠海角大学毕业证成绩单
一比一原版(CBU毕业证)不列颠海角大学毕业证成绩单
 
Chatty Kathy - UNC Bootcamp Final Project Presentation - Final Version - 5.23...
Chatty Kathy - UNC Bootcamp Final Project Presentation - Final Version - 5.23...Chatty Kathy - UNC Bootcamp Final Project Presentation - Final Version - 5.23...
Chatty Kathy - UNC Bootcamp Final Project Presentation - Final Version - 5.23...
 
一比一原版(UVic毕业证)维多利亚大学毕业证成绩单
一比一原版(UVic毕业证)维多利亚大学毕业证成绩单一比一原版(UVic毕业证)维多利亚大学毕业证成绩单
一比一原版(UVic毕业证)维多利亚大学毕业证成绩单
 
原版制作(Deakin毕业证书)迪肯大学毕业证学位证一模一样
原版制作(Deakin毕业证书)迪肯大学毕业证学位证一模一样原版制作(Deakin毕业证书)迪肯大学毕业证学位证一模一样
原版制作(Deakin毕业证书)迪肯大学毕业证学位证一模一样
 
社内勉強会資料_LLM Agents                              .
社内勉強会資料_LLM Agents                              .社内勉強会資料_LLM Agents                              .
社内勉強会資料_LLM Agents                              .
 
FP Growth Algorithm and its Applications
FP Growth Algorithm and its ApplicationsFP Growth Algorithm and its Applications
FP Growth Algorithm and its Applications
 
一比一原版(CU毕业证)卡尔顿大学毕业证成绩单
一比一原版(CU毕业证)卡尔顿大学毕业证成绩单一比一原版(CU毕业证)卡尔顿大学毕业证成绩单
一比一原版(CU毕业证)卡尔顿大学毕业证成绩单
 
一比一原版(RUG毕业证)格罗宁根大学毕业证成绩单
一比一原版(RUG毕业证)格罗宁根大学毕业证成绩单一比一原版(RUG毕业证)格罗宁根大学毕业证成绩单
一比一原版(RUG毕业证)格罗宁根大学毕业证成绩单
 
一比一原版(UofS毕业证书)萨省大学毕业证如何办理
一比一原版(UofS毕业证书)萨省大学毕业证如何办理一比一原版(UofS毕业证书)萨省大学毕业证如何办理
一比一原版(UofS毕业证书)萨省大学毕业证如何办理
 
Empowering Data Analytics Ecosystem.pptx
Empowering Data Analytics Ecosystem.pptxEmpowering Data Analytics Ecosystem.pptx
Empowering Data Analytics Ecosystem.pptx
 
一比一原版(Adelaide毕业证书)阿德莱德大学毕业证如何办理
一比一原版(Adelaide毕业证书)阿德莱德大学毕业证如何办理一比一原版(Adelaide毕业证书)阿德莱德大学毕业证如何办理
一比一原版(Adelaide毕业证书)阿德莱德大学毕业证如何办理
 
一比一原版(UPenn毕业证)宾夕法尼亚大学毕业证成绩单
一比一原版(UPenn毕业证)宾夕法尼亚大学毕业证成绩单一比一原版(UPenn毕业证)宾夕法尼亚大学毕业证成绩单
一比一原版(UPenn毕业证)宾夕法尼亚大学毕业证成绩单
 
一比一原版(CBU毕业证)卡普顿大学毕业证如何办理
一比一原版(CBU毕业证)卡普顿大学毕业证如何办理一比一原版(CBU毕业证)卡普顿大学毕业证如何办理
一比一原版(CBU毕业证)卡普顿大学毕业证如何办理
 
一比一原版(TWU毕业证)西三一大学毕业证成绩单
一比一原版(TWU毕业证)西三一大学毕业证成绩单一比一原版(TWU毕业证)西三一大学毕业证成绩单
一比一原版(TWU毕业证)西三一大学毕业证成绩单
 
Algorithmic optimizations for Dynamic Levelwise PageRank (from STICD) : SHORT...
Algorithmic optimizations for Dynamic Levelwise PageRank (from STICD) : SHORT...Algorithmic optimizations for Dynamic Levelwise PageRank (from STICD) : SHORT...
Algorithmic optimizations for Dynamic Levelwise PageRank (from STICD) : SHORT...
 
Ch03-Managing the Object-Oriented Information Systems Project a.pdf
Ch03-Managing the Object-Oriented Information Systems Project a.pdfCh03-Managing the Object-Oriented Information Systems Project a.pdf
Ch03-Managing the Object-Oriented Information Systems Project a.pdf
 

Profile injection attack detection in recommender system

  • 1. Detection of Profile Injection Attacks in Recommender System Presented By: Ashish Pannu
  • 2. Agenda • Recommender System - Introduction - Types of Recommender System - Why using Recommender System • Profile Injection Attacks - Why Attacker Attacks on Recommender System (Example) - Attack Profile Structure - Types of Attacks o Push Attacks o Nuke Attacks - Attack Detection Attributes o Generic Attributes o Model Specific Attributes • Proposed Approach • Conclusion • References 2
  • 3. Recommender System Introduction: – Recommender System (RS) is based on information filter to predict the rating or preference that a user would give to an unseen item [1]. – It is backbone of ecommernce websites (amazon, flipkart, myntra etc.), social networking websites (facebook, linkedin, twitter, google + etc.), matrimonial websites (shaadi.com, bharatmatrimonial.com etc.) and many more. 3
  • 4. Types of RS- 1. Collaborative Filtering (CF) based RS: based on correlation between different users. It states that if two user had similar tastes in past will also have same taste in future also. – Most popular and widely used. 2. Content based RS: based on the information on the features(keywords) of items rather than on the opinion of other users. E.g. in movie recommendation keywords may be: movie name, actor, actress, genre etc. 3. Knowledge based RS: customer defines his requirement explicitly. E.g. “I want black color BMW car”. Recommender System 4
  • 5. Recommender System – why? 1. Value for the customer – Narrow down the set of choice. – Discover new things. – Just exploring new items. – Make shopping easier. 2. Value for the provider – Increase user satisfaction. – Increase the sell. – Unique and personalized service to each customer. – Obtain more knowledge about the customer. – Opportunities for promotions. 5
  • 6. Profile Injection Attack Profile injection attack: Example • Assume that user-user based CF is used. • Pearson correlation as similarity measure. • Neighborhood size of 1. Item1 Item2 Item3 Item4 … Target Pearson Ram 5 3 4 1 … ? User1 3 1 2 5 … 5 -0.54 User2 4 3 3 3 … 2 0.68 User3 3 3 1 5 … 4 -0.72 User4 1 5 5 2 … 1 -0.02 6
  • 7. Profile Injection Attack Profile injection attack: Example • Assume that user-user based CF is used. • Pearson correlation as similarity measure. • Neighborhood size of 1. Item1 Item2 Item3 Item4 … Target Pearson Ram 5 3 4 1 … ? User1 3 1 2 5 … 5 -0.54 User2 4 3 3 3 … 2 0.68 User3 3 3 1 5 … 4 -0.72 User4 1 5 5 2 … 1 -0.02 User2 most similar to Ram 7
  • 8. Profile Injection Attack Profile injection attack: Example • Assume that user-user based CF is used. • Pearson correlation as similarity measure. • Neighborhood size of 1. User2 most similar to Ram Item1 Item2 Item3 Item4 … Target Pearson Ram 5 3 4 1 … ? User1 3 1 2 5 … 5 -0.54 User2 4 3 3 3 … 2 0.68 User3 3 3 1 5 … 4 -0.72 User4 1 5 5 2 … 1 -0.02 Attack 5 3 4 3 … 5 0.87 Attack 8
  • 9. Profile Injection Attack Profile injection attack: Example • Assume that user-user based CF is used. • Pearson correlation as similarity measure. • Neighborhood size of 1. Item1 Item2 Item3 Item4 … Target Pearson Ram 5 3 4 1 … ? User1 3 1 2 5 … 5 -0.54 User2 4 3 3 3 … 2 0.68 User3 3 3 1 5 … 4 -0.72 User4 1 5 5 2 … 1 -0.02 Attack 5 3 4 3 … 5 0.87 Attack Attack most similar to Ram 9
  • 10. Attack Profile Structure In order to look similar to genuine user, attacker gives ratings in a specific manner so that it become hard to identify the attack profile [2]. Selected items: these items are chosen because of their association with the target item. Filler items: these items are randomly chosen and rating is given based upon properties of attack. Unrated items: No ratings are given to these items. Target item: singleton item to which attacker promote or demote. Item1 … ItemK … ItemN … ItemR Target r_1 … r_k … r_l … r_n X Selected items Filler items Unrated items Target item 10
  • 11. Types of Attacks Push Attacks- to promote a specific item. Maximum rating will be given to the target item [3]. 1. Random Attack: – No rating to selected items. – Average rating of system to the filler items. – Least cost attack. – Limited effect as compared to more advance attacks. 2. Average Attack: – No rating to selected items. – Average rating of item to the item of filler items. – Additional cost of finding the average rating is involved. – More effective as compared to random attack. cont … 11
  • 12. Types of Attacks 3. Bandwagon Attack: – Maximum rating is given to popular items (selected items). – Average rating of system to the filler items. – It is a low cost attack. 4. Segment Attack: – Maximum rating is given to the items of the same segment of target item (selected items). – Minimum rating is given to filler items. – Best attack (impact wise) as compared to other attacks of same category. cont … 12
  • 13. Types of Attacks Nuke Attack- to demotion of an item. Minimum rating is given to the target item. 1. Love/Hate Attack: – No rating to selected items. – Maximum rating to set of filler items. 2. Reverse Bandwagon Attack: – Minimum rating is given to least popular items (selected item). – Average rating of system is given to the set of filler items. – More effective as compared to the love/hate attack. We can not stop attacks, we can just increase the cost of attacks. 13
  • 14. Attack Detection Attributes Generic Attributes- based on general abnormal behavior of user [4]. It is common for all attack types. 1. Rating Deviation from Mean Agreement (RDMA): finds profile’s average rating deviation per item. is the number of ratings given by user u. is rating given by user u to item i. is average rating of item i. is number of ratings provided for item i. 2. Degree Similarity with Top Neighbors (DegSim): find the average similarity of profile with top k neighbors. Nu r i,u ri ti 14
  • 15. Attack Detection Attributes 3. Length Variance: finds the variance in the length (# ratings given) of a given profile from the average length of database. Model Specific Attributes- focuses on the signature of attacks type. 1. Mean Variance (MeanVar): used for average and random attack. is the set of items rated by user u. 2. Filler Mean Target Difference (FMTD): used for bandwagon, reverse bandwagon and segment attack. is the set of filler items rated by user u. pu p f,u 15
  • 16. Evaluation Metrics 1. Precision: count total number of profiles that are labeled as attack [4]. 2. Recall: count total number of actual attack in the system. Reality Prediction Actually Attack Actually Good Rated Attack True Positive (Tp) False Positive (Fp) Rated Good False Negative (Fn) True Negative (Tn) 16
  • 17. Proposed Approach Steps- 1. Calculate the attack detection attributes (generic and model specific attributes). 2. Apply statistical models using k-fold cross validation 3. Compare the accuracy of statistical models. 4. Pick the top three performing models. 5. Ensemble the models (resulting models from step 4) using voting approach. Data Set Used: MovieLens-100K – 943 users – 1682 movies – 100000 ratings 17
  • 18. Performance Analysis Table 1: Performance analysis of 10% average attack. Table 2: Performance analysis for bandwagon attack at 5% filler size. Filler Size 1% 10% 20% 30% 40% 50% Models P R P R P R P R P R P R Decision Tree .90 .892 .921 .93 .928 .919 .939 .912 .94 .94 .961 .968 Random Forest .929 .930 .939 .92 .948 .948 .952 .956 .962 .961 .973 .979 Ada Boost .9 .908 .914 .918 .93 .924 .935 .938 .948 .934 .950 .943 SVM .938 .927 .949 .94 .959 .959 .971 .975 .979 .981 .988 .989 Linear Regression .89 .862 .89 .907 .918 .91 .925 .918 .929 .902 .93 .931 Neural Network .949 .938 .95 .943 .951 .954 .959 .958 .968 .967 .979 .98 Ensemble .939 .932 .946 .934 .953 .953 .961 .963 .968 .969 .98 .982 Attack Size 1% 3% 6% 9% 12% 15% Models P R P R P R P R P R P R Decision Tree .822 .811 .844 .830 .859 .848 .868 .855 .889 .872 .901 .928 Random Forest .880 .872 .894 .897 .905 .912 .909 .902 .929 .917 .957 .922 Ada Boost .819 .802 .829 .810 .834 .828 .858 .840 .883 .868 .902 .915 SVM .901 .902 .912 .928 .928 .908 .937 .911 .972 .959 .99 .984 Linear Regression .862 .842 .872 .882 .882 .908 .908 .919 .918 .935 .939 .941 Neural Network .892 .901 .919 .911 .922 .929 .959 .958 .961 .969 .97 .964 Ensemble .891 .892 .908 .912 .918 .916 .935 .924 .954 .948 .972 .957 18
  • 19. Conclusion 1. I tried to explain why RS is so important now a days for ecommerce. 2. I also tried to explain, how a attacker can manipulate the results of RS. 3. I focused on several keys areas of attacks against recommender system i.e. different types of attacks, attack detection attributes and model evaluation metrics. 4. I found that random forest, SVM and neural networks perform better than other models in attack detection. 5. I present ensemble approach for the attack detection and I found that performance is not best in any case but it also does not give worst performance in any case. 19
  • 20. 20
  • 21. References [1]. Davoodi, Fatemeh Ghiyafeh, and Omid Fatemi. "Tag based recommender system for social bookmarking sites." In Advances in Social Networks Analysis and Mining (ASONAM), 2012 IEEE/ACM International Conference on, pp. 934-940. IEEE, 2012. [2]. Lam, Shyong K., and John Riedl. "Shilling recommender systems for fun and profit." In Proceedings of the 13th international conference on World Wide Web, pp. 393-402. ACM, 2004. [3]. Mobasher, Bamshad, Robin Burke, Runa Bhaumik, and Chad Williams. "Toward trustworthy recommender systems: An analysis of attack models and algorithm robustness." ACM Transactions on Internet Technology (TOIT) 7, no. 4 (2007): 23. [4]. O'Mahony, Michael, Neil Hurley, Nicholas Kushmerick, and Guénolé Silvestre. "Collaborative recommendation: A robustness analysis." ACM Transactions on Internet Technology (TOIT) 4, no. 4 (2004): 344-377. 21