In　this　paper,　we　explore　the　problem　of　bandit　learning　based　dynamic　spectrum　anti-jamming　(DSAJ)　strategy　for　enabling　reliable　information　transmission　in　software　defined　UAV　swarm　(SD-UAS)　network.　We　first　present　UAV　swarm　network　architecture　based　on　software-defined　networking　(SDN)　and　describe　the　multiuser　multi-armed　bandit　(MAB)　based　anti-jamming　channel　selection　model　for　SD-UAS　network.　Then,　we　propose　a　collision　avoidance　(CA)　kl-UCB++　channel　selection　strategy　based　on　the　prior　spectrum　sensing　channel　information.　Simulation　results　validate　that　the　cumulative　regret　and　the　collisions　on　available　channels　obtained　with　the　proposed　spectrum　channel　selection　strategy　outperform　the　well-known　UCB,　kl-UCB++　learning　algorithm,　which　can　effectively　improve　the　anti-jamming　ability　of　SD-UAV　swarm　in　complicated　electromagnetic　jamming　environment.　©　2020　IEEE.