Papers › Provable defenses against adversarial examples via the convex outer adversarial polytope

Provable defenses against adversarial examples via the convex outer adversarial polytope

2 Nov 2017ICML 2018 7arXiv:1711.00851archive 2025-07-28

Eric Wong, J. Zico Kolter

We propose a method to learn deep ReLU-based classifiers that are provably robust against norm-bounded adversarial perturbations on the training data. For previously unseen examples, the approach is guaranteed to detect all adversarial examples, though it may flag some non-adversarial examples as well. The basic idea is to consider a convex outer approximation of the set of activations reachable through a norm-bounded perturbation, and we develop a robust optimization procedure that minimizes the worst case loss over this outer region (via a linear program). Crucially, we show that the dual problem to this linear program can be represented itself as a deep network similar to the backpropagation network, leading to very efficient optimization approaches that produce guaranteed bounds on the robust loss. The end result is that by executing a few more forward and backward passes through a slightly modified version of the original network (though possibly with much larger batch sizes), we can learn a classifier that is provably robust to any norm-bounded adversarial attack. We illustrate the approach on a number of tasks to train classifiers with robust adversarial guarantees (e.g. for MNIST, we produce a convolutional classifier that provably has less than 5.8% test error for any adversarial attack with bounded ℓ_∞ norm less than ϵ= 0.1), and code for all experiments in the paper is available at https://github.com/locuslab/convex_adversarial.

PaperPDFConference PDFCodeCode Syntology ran

In Syntology View this paper on Syntology: its repositories, every harvested function with whether it ran, its licence and the call to fetch it.

Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

For agents, Syntology's MCP tool lists every function and class Syntology harvested from this paper and whether it ran (how to connect): get_harvested_code_for_paper(arxiv_id="1711.00851")

Code

Syntology Ran 21 of 44 code samples harvested from 5 repositories linked to this paper; 23 have no recorded run. Of those that ran: 1 ran · honoured contract; 1 ran · violated contract; 1 ran · our draft was wrong; 1 ran · fixture could not drive it; 17 ran with no contract checked.

By repository: official repository: 6 samples from 1 repository, 4 ran; community (archive-listed): 38 samples from 4 repositories, 17 ran. The run record, sample by sample. “Ran” means executed on a synthesized input, not that the code is correct or reproduces the paper.

locuslab/convex_adversarial officialmentioned in papermentioned on GitHubpytorch report
Hadisalman/robust-verify-benchmark mentioned on GitHubpytorch report
ermongroup/generative_adversary mentioned on GitHubtf report
fra31/mmr-universal mentioned on GitHubpytorch report
vtjeng/MIPVerify.jl mentioned on GitHub report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

44 samples harvested; 21 ran; 1 honoured the contract we drafted; 23 have no recorded run. Read from Syntology's graph 2026-09-25; that is when this build read the record, not when the samples ran.

1ran · honoured contract
1ran · violated contract
1ran · our draft was wrong
1ran · fixture could not drive it
17ran
23unverified

Licence: 2 of the 44 samples are pointer only, meaning Syntology does not serve that copy's text. This page shows no code text for any sample; each one links to its file in the repository.

Harvested from 5 repositories linked to this paper, official or community; each sample names its own and says which. “Ran” means the sample executed on a synthesized input. It does not mean the output is correct, and nothing here reproduces the paper's results. “Honoured” and “violated” refer to a contract Syntology drafted from the code itself; “our draft was wrong” and “fixture could not drive it” are failures of Syntology's instrument, not of the code.

Each sample ends with its code_sha256, Syntology's identity for that exact code. An agent fetches the stored sample with Syntology's MCP tool get_code(code_sha256="…") (how to connect); click an identity to copy that call.

Repository labels, per sample. official repository: The archive marks this repository official for the paper. named in the paper: The archive records that the paper mentions this repository; it is not marked official. community (archive-listed): In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper. found in paper text by Syntology: Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted. community: Not in the archive's code links for this paper; a community repository Syntology harvested. Samples from a repository marked official are listed first. Licence labels name the repository's licence as recorded at harvest. “Pointer only” means Syntology does not serve that copy's text, for one of four reasons: no licence file was found; the licence was not identified; the licence is recorded as permissive but that copy's record is not marked cleared; or the licence is outside the permissive list Syntology serves text under (MIT, Apache-2.0, BSD and similar). Some licences outside that list permit redistribution, such as WTFPL, and GPL-3.0 under its conditions; they are simply not on the list. Hover a licence label for the reason. File links open the file on GitHub at the default branch, which may have changed since the harvest.

InfBall locuslab/convex_adversarial/convex_adversarial/dual_network.py official repository ran MIT (permissive) · a31183fb505bf309 · report
InfBallBounded locuslab/convex_adversarial/convex_adversarial/dual_network.py official repository ran MIT (permissive) · 37ea0914d3143c9f · report
InfBallProj locuslab/convex_adversarial/convex_adversarial/dual_network.py official repository ran fingerprinted MIT (permissive) · 49cb906dccd76915 · report
InfBallProjBounded locuslab/convex_adversarial/convex_adversarial/dual_network.py official repository ran MIT (permissive) · 5a79e42b5ecc22cd · report
DualNetwork locuslab/convex_adversarial/convex_adversarial/dual_network.py official repository unverified MIT (permissive) · 34e633f7b59f06cd · report
select_input locuslab/convex_adversarial/convex_adversarial/dual_network.py official repository unverified MIT (permissive) · 5187a83994dc73d4 · report
Conv2dUntiedBias fra31/mmr-universal/kolter_wong/convex_adversarial/dual_network.py community (archive-listed) ran · metamorphic tier: deterministic MIT (permissive) · 9634a2860dbe32ed · report
DualLinear fra31/mmr-universal/kolter_wong/convex_adversarial/dual_network.py community (archive-listed) ran · metamorphic tier: deterministic MIT (permissive) · a465ad16a8510751 · report
DualReLU fra31/mmr-universal/kolter_wong/convex_adversarial/dual_network.py community (archive-listed) ran · metamorphic tier: deterministic MIT (permissive) · 4d488a7c833e637b · report
DualReLUProj fra31/mmr-universal/kolter_wong/convex_adversarial/dual_network.py community (archive-listed) ran · metamorphic tier: deterministic MIT (permissive) · 733e1d54c3d21ea9 · report
DualReLUProj Hadisalman/robust-verify-benchmark/wong_kolter/dual_network.py community (archive-listed) ran · metamorphic tier: deterministic MIT (permissive) · a721c225e4045355 · report
DualReshape fra31/mmr-universal/kolter_wong/convex_adversarial/dual_network.py community (archive-listed) ran · metamorphic tier: deterministic MIT (permissive) · d1e8d479da8d224c · report
Identity fra31/mmr-universal/kolter_wong/convex_adversarial/dual_network.py community (archive-listed) ran · metamorphic tier: deterministic MIT (permissive) · 15451c00600189d1 · report
InfBall Hadisalman/robust-verify-benchmark/wong_kolter/dual_network.py community (archive-listed) ran · metamorphic tier: deterministic MIT (permissive) · 7b3dd3557f3b683e · report
InfBallBounded fra31/mmr-universal/kolter_wong/convex_adversarial/dual_network.py community (archive-listed) ran · metamorphic tier: deterministic MIT (permissive) · 0607bcee1940ecc6 · report
InfBallProj Hadisalman/robust-verify-benchmark/wong_kolter/dual_network.py community (archive-listed) ran MIT (permissive) · d0f7cdcc96ab6561 · report
InfBallProjBounded fra31/mmr-universal/kolter_wong/convex_adversarial/dual_network.py community (archive-listed) ran MIT (permissive) · 9568161abd5fcecf · report
InfBallProjBounded Hadisalman/robust-verify-benchmark/wong_kolter/dual_network.py community (archive-listed) ran MIT (permissive) · 341ef71271a4cbef · report
L2Ball Hadisalman/robust-verify-benchmark/wong_kolter/dual_network.py community (archive-listed) ran · metamorphic tier: deterministic MIT (permissive) · 006a937f9cba3174 · report
batch fra31/mmr-universal/kolter_wong/convex_adversarial/dual_network.py community (archive-listed) ran · our draft was wrong fingerprinted MIT (permissive) · 778274b3fb463e7e · report
full_bias fra31/mmr-universal/kolter_wong/convex_adversarial/dual_network.py community (archive-listed) ran · honoured contract MIT (permissive) · 2a4b1b5d869483b9 · report
get_layer_bound_relax_adaptive_matrix_huan_optimized huanzhang12/CROWN-Robustness-Certification/get_bounds_ours.py community (archive-listed) ran · fixture could not drive it licence not identified · pointer only · 728154bbcd429f2b · report
unbatch fra31/mmr-universal/kolter_wong/convex_adversarial/dual_network.py community (archive-listed) ran · violated contract fingerprinted MIT (permissive) · 9efb427bc4cdf3df · report
AditiMNIST ermongroup/generative_adversary/models/aditi_mnist.py community (archive-listed) unverified GPL-3.0 (copyleft) · pointer only · f301dfa41009abd3 · report
Dense fra31/mmr-universal/kolter_wong/convex_adversarial/dual_network.py community (archive-listed) unverified MIT (permissive) · bc7271d7a6811f02 · report
DenseSequential fra31/mmr-universal/kolter_wong/convex_adversarial/dual_network.py community (archive-listed) unverified MIT (permissive) · 18133b07da9a458e · report
DualBatchNorm2d fra31/mmr-universal/kolter_wong/convex_adversarial/dual_network.py community (archive-listed) unverified MIT (permissive) · c52d8aff384b08a4 · report
DualConv2d fra31/mmr-universal/kolter_wong/convex_adversarial/dual_network.py community (archive-listed) unverified MIT (permissive) · b23673a5bc8ae4a6 · report
DualConv2d Hadisalman/robust-verify-benchmark/wong_kolter/dual_network.py community (archive-listed) unverified MIT (permissive) · 3d025a7e72e695af · report
DualDense fra31/mmr-universal/kolter_wong/convex_adversarial/dual_network.py community (archive-listed) unverified MIT (permissive) · d11e9d2b6bc1c8a7 · report
DualDense Hadisalman/robust-verify-benchmark/wong_kolter/dual_network.py community (archive-listed) unverified MIT (permissive) · 5428007ec654a464 · report
DualLayer fra31/mmr-universal/kolter_wong/convex_adversarial/dual_network.py community (archive-listed) unverified MIT (permissive) · 46dfb83f015faad4 · report
DualNetwork fra31/mmr-universal/kolter_wong/convex_adversarial/dual_network.py community (archive-listed) unverified MIT (permissive) · 2f26c3676bd1d6ca · report
DualNetwork Hadisalman/robust-verify-benchmark/wong_kolter/dual_network.py community (archive-listed) unverified MIT (permissive) · d09a9fa8766395b6 · report
DualObject fra31/mmr-universal/kolter_wong/convex_adversarial/dual_network.py community (archive-listed) unverified MIT (permissive) · 7e8fbd57df2c528e · report
InfBall fra31/mmr-universal/kolter_wong/convex_adversarial/dual_network.py community (archive-listed) unverified MIT (permissive) · 0b1a6f38f8307d1f · report
InfBallProj fra31/mmr-universal/kolter_wong/convex_adversarial/dual_network.py community (archive-listed) unverified MIT (permissive) · edc9bd1055691c58 · report
L2BallProj Hadisalman/robust-verify-benchmark/wong_kolter/dual_network.py community (archive-listed) unverified MIT (permissive) · 610e37ee41f7c916 · report
conv2d fra31/mmr-universal/kolter_wong/convex_adversarial/dual_network.py community (archive-listed) unverified MIT (permissive) · 26c63c3b5b5c6239 · report
conv_transpose2d fra31/mmr-universal/kolter_wong/convex_adversarial/dual_network.py community (archive-listed) unverified MIT (permissive) · 73dd3a4a0039021b · report
select_input fra31/mmr-universal/kolter_wong/convex_adversarial/dual_network.py community (archive-listed) unverified MIT (permissive) · 4fb58a8213bd4638 · report
select_input Hadisalman/robust-verify-benchmark/wong_kolter/dual_network.py community (archive-listed) unverified MIT (permissive) · 935601b807dbd64e · report
select_layer fra31/mmr-universal/kolter_wong/convex_adversarial/dual_network.py community (archive-listed) unverified MIT (permissive) · 822d6e0702879c11 · report
select_layer Hadisalman/robust-verify-benchmark/wong_kolter/dual_network.py community (archive-listed) unverified MIT (permissive) · 96d6186dbe64fc7e · report

Tasks

Adversarial Attack

Results from the paper archive 2025-07-28

No leaderboard rows for this paper in the archive.

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections