In this paper, the influence of the social media on the opinion formation process is modeled during an election campaign. In the proposed model, peer- to-peer interactions and targeted online propaganda messages are assumed to be the driving forces of the opinion formation dynamics. The conviction power of the targeted messages is based on the collected and processed private information. In this work, the model is based on an artificial society, initially evenly divided between two parties. The bounded confidence model governs peer-to-peer interactions with a value of confidence parameter which leads to consensus. The targeted messages which was modeled as an external interacting source of information convert some weakly committed individuals to break this evenness. Both parties use the same methods for propaganda. It is shown that a very small external influence break the evenness of the opinion distribution which play significant role in the election results. Obtained opinion fluctuation time series have close resemblance with the actual election poll results.