Apart from well known malwares viruses, worms and Trojan houses; there is less familiar threat known as the botnet. The term botnet (network of bots) is a combination of two words: bot (victim host) and net (network). In relation of botnet taxonomy bot is referred as a victim host which is under the control of the attacker called BotMaster (or Botherder). These botnets are frequently used for many cyber attacks and crimes, and they are root causes for several illegal activities like click fraud, DDOS, etc. Botnets operate under the command and control infrastructure (C & C) which makes botnets functioning unique giving serious problems in defending from this malware. Botnets become more elaborate and efficient. Their use is growing at an exponential rate. Although botnets showed their existence several years ago, it became an interesting area for research only recently. Various types of technique are proposed for detection and prevention from botnet attacks. Current detection models deal with only a limited set of bots behavior and thus are not able to resolve protocol independent and architecture independent (PI & AI) problem, and autoupdation mechanism used by the botnet. The proposed model addresses these problems along with the detection of advanced botnets. In this paper, we have taken up a survey report for detection of hybrid botnets.