
Latest from our lab and first paper of 2024🔥
Publication alert!🚀Our paper "Mechanisms Driving Online Vaccine Debate During the COVID-19 Pandemic" which we have started in @SICSS_Istanbul in 2022 and @teyitorg is published in Social Media+Society It was a great journey with @onurvarol @aybukeatalayy @egeotenen @umut_duyguu
Our method has implications in the study of coordinated campaigns on social media. Furthermore, this method is applicable to any social media platform in which an ordered list of followers and their profile creation dates are attainable.
We released a tool to analyze the followers of a social media account “FollowerAnalyzer”. It incorporates our fake follower detection algorithm, follow time estimation, and several plotting utilities.
github.com/ViralLab/Follower…
6
323
We hypothesize three possible reasons for a user to be followed by coordinated fake followers:
(i) The user purchases followers to manipulate engagement metrics
(ii) Fake accounts target the user to serve propaganda
(iii) Fake accounts follow users to fit a specific persona
1
3
215
We also examined the behavior of coordinated fake followers across different users by identifying a group of fake followers and searching for them in the followers of the remaining 1300 Twitter users. We found that these anomalous accounts also follow other users simultaneously.
1
3
204
We are excited to share our findings in the recent preprint by @yasser_zouzou, which provides a novel method for detecting coordinated fake followers and investigates their prevalence in Turkish political Twitter
Paper: arxiv.org/abs/2310.20407
Code: github.com/ViralLab/Follower…
We found that the detected coordinated fake followers are not always captured by other bot detectors. Additionally, we observed that the number of followers and shared statuses vary significantly between fake followers and normal followers.
1
2
168
Using the Secim2023 dataset published by VRL Lab (secim202.biz), we uncovered groups of fake followers of 1300 Twitter accounts belonging to Turkish politicians and media outlets.
1
2
146
We used a dataset from the Dribbble platform with inserted synthetic fake followers to test our method and compare it to existing anomaly detection methods, which we applied using features engineered from the follower map. Sliding Histogram clearly outperformed the other methods
1
1
140
Our suggested method, “Sliding Histogram”, uses a sliding window to compare the density of followers in different parts of the follower map. Followers in zones that have an anomalously high density are assigned higher anomaly scores.
1
1
142
The coordinated fake followers that we target can be distinguished by their anomalous follow patterns in a “follower map”. A follower map is a graph that plots all the followers of a social media account based on their follow rank and profile creation dates.
1
1
162
The coordinated fake followers we detect are groups of accounts that simultaneously follow a user on social media. This type of fake followers was observed in previous studies, like this investigation by the New York Times: nytimes.com/interactive/2018…
1
1
234
New paper alert ⏰
We built and publicly released TurkishBERTweet for social media analysis.
Pre-print: arxiv.org/abs/2311.18063
HuggingFace: huggingface.co/VRLLab/Turkis…
Considering the low-inferance time and public availablity, this model is suitable for large-scale Turkish social media analysis.
1
179
Seçimin ikinci tura kalmasıyla birlikte en çok merak edilen sorulardan biri ilk turda Sinan Oğan'a oy veren %5.2'lik seçmenin davranışı oluyor.
Yaptığımız analizde siyasilerin takipçi paylaşımları ve ne seviyede ortaklıkları olduğunu inceledik.
secim2023.biz/posts/230520/
Yarın Cumhuriyet tarihinin en önemli seçimi icin sandıklara oy kullanmaya gideceğiz 🗳️ Adayları, liderleri ve partileri sosyal medya perspektifinden bir kez daha incelemeye ne dersiniz?
⬇️
secim2023.biz/apps/politicia…
Geliştirdiğimiz siyasi profilleri inceleme arayüzü ile son aylarda siyasilerin attığı tweetleri aldığı etkileşimleri, takipçi sayılarındaki değişimi ve güncel takipçilerinin yaş ve cinsiyet dağılımlarını görebilirsiniz.
1
4
516















