Backend
Propose with the Gradient, Accept with the Loss: One Hedged Attack for Pixel-Sparse Robustness
arXiv cs.CR0 views
arXiv:2611.22596v1 Announce Type: cross Abstract: A pixel-sparse adversary may change a few pixels of an image to any value. Robustness to such attacks is currently measured with an ensemble, sparse-AutoAttack (sAA), because its members fail in different places: sparse projected gradient descent (sPGD) trusts a first-order model that a full pixel change breaks, and sparse random search (Sparse-RS) ignores the gradient even where it is informative. We show that both attacks are local searches, each a proposal distribution plus an acceptance rule, and ask what such a search can learn from the gradient. A Taylor bound says what to measure, and measurements at the states the search visits give five rules. The sign of the gradient is its most reliable use at a single pixel and sets a new pixel's value. Its ranking of candidate pixels should be followed loosely, and its ranking of the pixels to remove, which is often more reliable, can be followed more closely. A pixel should not be proposed twice before a move is accepted, and acceptance should be decided by the loss alone. These rules define Gradient-guided Random Search (GRS), which runs once and costs one forward and one backward pass per step. On 13 CIFAR-10 configurations and 5 ImageNet models, GRS matches or beats the best single member of sAA and is within one point of sAA itself or stronger on 17 of 18 settings, and it is significantly stronger than sAA on 8. It uses 3 instead of up to 7 forward-pass units, and on CIFAR-10 it runs 3.7 to 5.0 times faster than the members of sAA run in full.
Source:https://arxiv.org/abs/2611.22596PreviousDexFormer 2.0: Towards a Cross-Embodiment Foundation Policy for Dexterous GraspingNextHE-LRM: Encrypted Deep Learning Recommendation Models using Fully Homomorphic Encryption
Related posts
Your analytics export says more than you think (if you read it honestly)
0dev.to OpenSource
Tencent LoopForge: Turn Coding Agents Into Repeatable Software Delivery
0dev.to OpenSource
Your agent does not need more tools. It needs a checkpoint.
0dev.to OpenSource
NVIDIA SkillEvaluator 1.0.1: Test Agent Skills Before Shipping Them
0dev.to OpenSource
I routed Claude Code tasks to cheaper models. Here's what 15 requests saved. published: true tags: claudecode, ai, opensource, productivity
0dev.to OpenSource
Who Owns the Code AI Wrote?
0dev.to Programming