anishathalye

10 ms·

31.
Compromising Garmin Sport Watches: A Deep Dive into GarminOS and Its MonkeyC VM(anvilsecure.com)
176 points by anishathalye 3y ago|74 comments
32.
CleanVision: Audit Your Image Data for Better Computer Vision(cleanlab.ai)
4 points by anishathalye 4y ago|0 comments
33.
by anishathalye 4y ago
The class website has a list of resources that includes open-source DCAI tools: https://dcai.csail.mit.edu/resources/
34.
by anishathalye 4y ago
Hi HN! I’m back with another “what they don’t teach you in school” style course that I’d love to share with the community. (A couple years ago, I was part of the team that taught Missing Semester, an IAP class that taught programmer tools t
35.
MIT Introduction to Data-Centric AI(dcai.csail.mit.edu)
4 points by anishathalye 4y ago|1 comments
36.
Tips for a Data-Centric AI Approach(landing.ai)
1 points by anishathalye 4y ago|0 comments
37.
by anishathalye 4y ago
Many real-world datasets use multiple annotations per example to ensure higher-quality labels. CROWDLAB is a new set of algorithms that estimate 3 key quantities better than prior standard crowdsourcing algorithms like GLAD and Dawid-Skene:
38.
Crowdlab: Effective algorithms to handle data labeled by multiple annotators(cleanlab.ai)
2 points by anishathalye 4y ago|1 comments
39.
by anishathalye 4y ago
We’re excited to share with the HN community our open-source Python package for addressing data-quality issues in machine learning, automating tasks like finding label errors in datasets. cleanlab started out as a grad student research proj
40.
Cleanlab 2.0: Automatically Find Errors in ML Datasets(cleanlab.ai)
2 points by anishathalye 4y ago|1 comments
41.
by anishathalye 5y ago
Yes, I try to do this, but it's extra effort to remember. I usually try to clean up my public repositories once every couple of months.
42.
by anishathalye 5y ago
I've thought about doing that, but I find the "Merge branch 'anishathalye-forked-repos/...'" that appears when people click the merge button on GitHub unattractive.
43.
by anishathalye 5y ago
I wish GitHub had a better way of distinguishing repositories (all the random experimental repos I've created or repos I've forked for the purpose of making a small contribution, sorted by last modified time) that show up on the p
44.
by anishathalye 5y ago
As others in this thread have pointed out, some of those proofs involve calls to high-power decision procedures, so I was curious to see if as a baseline, could Z3 solve the problems in the blog post with no guidance? For 5/6 of the pr
45.
by anishathalye 5y ago
Yes, that would have made sense, but I was being lazy :) I also didn't notice this issue until after I had trained the model and tried inverting hashes (didn't want to look through the NSFW dataset myself). And I thought it was am
46.
by anishathalye 5y ago
A bit of context: Microsoft developed PhotoDNA to identify illegal images like CSAM -- NCMEC maintains a database of PhotoDNA signatures, and many companies use this service to identify and remove these images. Microsoft claims: > A Phot
47.
Inverting PhotoDNA(anishathalye.com)
131 points by anishathalye 5y ago|25 comments
48.
by anishathalye 5y ago
Even though it's not apparent from this demo website, the underlying code performs a preimage attack (it doesn't actually use a target image, just a target hash), so it's even stronger than a second preimage attack.
49.
by anishathalye 5y ago
The README ( https://github.com/anishathalye/neural-hash-collider#how-it-... ) explains in a bit more detail how these adversarial attacks work. This code pretty much implements a standard adversarial attack against Neur
50.
Show HN: Neural-hash-collider – Find target hash collisions for NeuralHash(github.com)
623 points by anishathalye 5y ago|351 comments
51.
by anishathalye 6y ago
Also of interest might be cleanlab ( https://github.com/cgnorthcutt/cleanlab ), the open-source software we used for initially identifying potentially mislabeled data.
52.
Show HN: Label errors in benchmark ML test sets(labelerrors.com)
2 points by anishathalye 6y ago|2 comments
53.
by anishathalye 6y ago
Hi HN! One of the authors here. We found pervasive errors in the test sets of 10 of the most commonly used benchmark ML datasets, so we made labelerrors.com where anyone can examine the data labels. We think it’s neat to browse through the
54.
by anishathalye 6y ago
Hi HN! Recently, I decided to take care of a task I had been procrastinating for a while: to organize and de-dupe data on our home file server. I was thinking of it as a mundane task that needed to get done at some point, but the problem tu
55.
Organizing Data Through the Lens of Deduplication(anishathalye.com)
4 points by anishathalye 6y ago|1 comments
56.
by anishathalye 7y ago
https://www.anishathalye.com/ I write mostly about playful hacks that I have worked on. Many posts tend to have a researchy flavor. Some of my favorites: * Turning a MacBook into a touchscreen with $1 of hardware: https:&#
57.
by anishathalye 7y ago
On macOS, we used this excellent open-source software called KeyCastr: https://github.com/keycastr/keycastr (we used this e.g. in the text editors lecture)
58.
The missing semester of CS education(missing.csail.mit.edu)
1163 points by anishathalye 7y ago|196 comments
59.
by anishathalye 7y ago
Thank you all for the pointers to related projects and research papers! To address a couple common questions: Q: Is Basalt open source? A: Not yet, but it will be. The language is still under heavy development, and it's currently being
60.
by anishathalye 7y ago
Cassowary is an incremental solver for linear systems of equations and inequalities, while Basalt's constraints are more expressive (which is why it needs an SMT solver). One example of something that's not expressible in a linear

More ›