Redeem Myself: Purifying Backdoors in Deep Learning Models using Self Attention Distillation | IEEE Conference Publication | IEEE Xplore