Social networks are recognized as popular communication channel but in this, there is one of the problem is spam messages. Spam messages can contain malware in the form of the executable file and the link to the malicious websites or the links which do not exist. Most of the existing machine learning solutions are based on the either Support Vector Machine or Naive Bayes but the existing solutions either slow or inaccurate in solving spam filtering problem. Support Vector Machine based spam filter has great advantages on high precision and recall rate and Naive Bayes based spam filter give faster classification speed and require small training sets. By taking the advantages of both, we propose hybrid spam filtering algorithm which have more accuracy than separately implemented NB and SVM.