Downstream Task-agnostic Transferable Attacks on Language-Image Pre-training Models | IEEE Conference Publication | IEEE Xplore