Spammedin: An AI-Powered Multi-Agent Framework for LinkedIn Scam Profile and Organization Detection

Professional-networking sites like LinkedIn have become central to hiring, career development, and business outreach, but their massive user base has also turned them into a favourite hunting ground for impersonated recruiters, sham companies, and misleading job or internship postings that prey on unsuspecting users. As the platform grows, reviewing every profile or company page by hand becomes impossible, which calls for a verification approach that is both automated and easy to interpret. This work introduces Spammedin, a multi-agent AI system that estimates the trustworthiness of a given LinkedIn profile or organization by pooling evidence gathered from text, domain records, visual media, and behavioural cues. Four dedicated agents — Profile, Organization, Image, and Behaviour — work in parallel to assess distinct credibility markers such as account age, how complete the profile information is, connection totals, endorsement trends, how long a domain has been registered, site reputation, image authenticity, and posting habits. Outputs from these agents are merged in a Feature Fusion stage and then evaluated by a Risk Assessment Agent, which produces a risk value ranging from 0 to 100 and tags the submission as Genuine, Suspicious, or Scam. An Explainability Agent applies SHAP and LIME on top of each verdict to generate a rationale that a human reader can follow. A working prototype was implemented and tested on sample profile and organization data, and the outcomes show that the multi-agent pipeline functions as intended.