Search papers, labs, and topics across Lattice.
University of Wisconsin鈥揗adison
3
0
6
23
Provably undetectable backdoors can be injected into pre-trained image classifiers, even with white-box access, by exploiting sparse perturbations and Gaussian dithering.
LLM-powered reverse engineering agents still stumble on basic obfuscation, showing we're further from automated vulnerability discovery than recent hype suggests.
Web agents are surprisingly bad at basic website security and privacy tasks, failing almost half the time when confronted with common UI elements like toggles and checkboxes.
