Shilling attack detection for recommender systems

2022 COE Engineering Design Project (RK01)

---

Faculty Lab Coordinator

Rasha Kashef

Topic Category

Software Systems

Preamble

In commerce, recommendation systems plays a great role of providing recommendation to users, however, recommendation systems are vulnerable to attacks. Attacks on recommender system behaviour is known as a “shilling” attack or “profile injection” attack. Users that carry out shilling attacks are known as attackers or Shillers. Advances in machine learning and context-awareness systems have provided a great scope of detecting such shillings attacks, namely the "shillings detector". Through predictive analytics of user-ratings and reviews, the "smart attack detector" can provide real-time detection of attacks which helps in manipulating the correct decision while providing accurate recommendations.

Objective

The objective of this project is to build a smart real-time shillings detectors that can provide automated detection of shillings attacks while providing recommendations

Partial Specifications

1- The data could be large which requires intensive Manipulation
2- converting contexts should be used by some NLP analysis
3- injecting the contexts into the shillings detector should have a specific data model

Suggested Approach

1. Analyze the collected data to detect shillings profiles using well-known AI-based shillings detectors
2. Inject contexts such as reviews to allow more robustness of the state-of-the art detection models
3- design a novel model that uses the contexts and machine learning
4. Play analysis results back through validation processes .

Group Responsibilities

1. Survey literature to figure out the best approach.
2. full shillings detector system design (pilot model)
3. Implementation (production model)
4. Testing under varying conditions (ratings, user reviews, multi-context, noisy data, etc.)
5- Complete Analysis and validation of the models
6- Results documentation and report writing

Student A Responsibilities

1- Data collection
2- data manipulation
3- data filtering
4- data standardization
5- data normalization

Student B Responsibilities

1- Programming for the state-of-the art recommendation systems
2- Programming for the state-of-the art shillings detector recommendation systems

Student C Responsibilities

1- Design the shillings detector
2- programming for the designed context-aware shillings detector

Student D Responsibilities

1- programming for the validation of shillings detector
2- reporting results and documentation

Course Co-requisites

To ALL EDP Students

Due to COVID-19 pandemic, in the event University is not open for in-class/in-lab activities during the Winter term, your EDP topic specifications, requirements, implementations, and assessment methods will be adjusted by your FLCs at their discretion.