Detection of Fake Profile in Online Social Networks Using Machine Learning

Information

Meta

Goals

Result

The amount of research work is very low to recognize counterfeit characters made by people

Recognize fake accounts

2018

Creating false account

Reasoning

Peoples and groups who forge their identities on social media platforms are striving to spread chaos in our society.

We are not expecting that people will give their accurate details in the privacy policies of social media platforms

Now a day’s fake accounts can be bought online at a very less cost and can be given to the customer via crowdsourcing services. Now a day it is easier to buy Twitter and Instagram followers and likes online

Alter the credibility of any account so that it can be used to spread false news, spreading rumours. Scandalize someone’s character polarising opinions

Related works

Techniques which have been applied or similar to detect, identify and eliminate fake accounts

If a certain message contains these words or number of words these are addressed to as spam.

Unsupervⅰsed ML

Main drawback is that the process of developing new words are easy and constant and the use of shortened words are becoming more common on platform for example lol

Facebook uses algorithms which can identify bots using the number of friends for deceptiveness which could either be related to relationships history or tagging.

Successful for identifying bot accounts by have not been successful in identifying fake accounts by humans

Administered ML

Requires dataset of highlights with a name arranging each column or result.

Hⅰghlⅰghts are therefore information utilized by regulated ML models to foresee a result. These highlights can be the characteristics discovered by means of APIs

Author

Naman Singh

Tushar Sharma

Abha Thakral

Tanupriya Choudhury

University

Amity University Uttar Pradesy

University of Petroleum and Energy Studies

Technique "Bunching"

Could be utilized for distinguised bot

Information not labeled, but are assembled in view of closeness

Traits example (Twitter)

name

screenname

created

friends_count

followers_count

language

language

listed_count

profile_image

status_count

location

timezone

utc_offset

Designed highlights differentiated to 3 gatherings

potraying the data of the accounts

the connection between record and others

record behavior and message

To recognize counterfeit fake bot accounts in SOCIAL MEDIA PLATFORMSs, there are combinations developed by connecting various engineered features to machine learning models

Cresci et al in their model shows that the characteristic features of the record is enough to distinguish the record between the normal or bots

Gupta et al. conducted and proposed, the recurrence and types of messages and on what particular time of day, gives more data pertinent to trickiness than the characteristics features of the record itself [11].