<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Privacy on Suriya's site</title><link>https://subygan.com/tags/privacy/</link><description>Recent content in Privacy on Suriya's site</description><generator>Hugo</generator><language>en</language><copyright>This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.</copyright><item><title>right to be forgotten</title><link>https://subygan.com/tech/privacy/right_to_forget/</link><pubDate>Tue, 06 Aug 2024 00:00:00 +0000</pubDate><guid>https://subygan.com/tech/privacy/right_to_forget/</guid><description>&lt;p>This is a (mostly) search engine specific requirement where individuals are protected from the harm of their past actions.&lt;/p>
&lt;h2 id="disparity-in-countries">disparity in countries&lt;/h2>
&lt;p>Similar to any privacy issue, these countries are divided in the way these laws are implemented. Especially between developed countries such as United States and the European Union.&lt;/p></description></item><item><title>privacy problems and attacks in machine learning</title><link>https://subygan.com/tech/privacy/privacy_ml/</link><pubDate>Fri, 19 Jul 2024 00:00:00 +0000</pubDate><guid>https://subygan.com/tech/privacy/privacy_ml/</guid><description>&lt;h2 id="privacy-problems-and-attacks-in-machine-learning">privacy problems and attacks in machine learning&lt;/h2>
&lt;ul>
&lt;li>&lt;strong>Membership inference attacks&lt;/strong>: These aim to determine if a particular data point was used to train a model. Researchers study how to detect and mitigate such attacks to protect individuals&amp;rsquo; privacy.&lt;/li>
&lt;li>&lt;strong>Model inversion attacks&lt;/strong>: These attempt to reconstruct training data from model outputs. Understanding these vulnerabilities helps in developing more robust privacy-preserving techniques.&lt;/li>
&lt;li>&lt;strong>Data reconstruction in federated learning&lt;/strong>: This examines how malicious participants could potentially extract private information from other participants during collaborative training.&lt;/li>
&lt;li>&lt;strong>Model extraction attacks&lt;/strong>: These involve extracting a model&amp;rsquo;s architecture or parameters, which could be used to launch further attacks or to steal intellectual property.&lt;/li>
&lt;li>&lt;strong>Data Poisoning in Edge ML&lt;/strong>: This is a problem where the data that is being used to train the model is poisoned by an attacker. This is a huge problem in edge ML where the data is not curated and is coming from a variety of sources.&lt;/li>
&lt;li>&lt;strong>Differential privacy techniques&lt;/strong>: Researchers explore how to add controlled noise to data or model outputs to preserve privacy while maintaining utility.&lt;/li>
&lt;li>&lt;strong>Secure multi-party computation&lt;/strong>: This involves developing protocols for multiple parties to jointly compute on private data without revealing individual inputs.
Homomorphic encryption: This allows computations on encrypted data, potentially enabling privacy-preserving machine learning on sensitive information.
Privacy-preserving record linkage: Techniques to securely combine datasets from multiple sources without compromising individual privacy.&lt;/li>
&lt;/ul>
&lt;h2 id="papers-and-other-readings">papers and other readings&lt;/h2>
&lt;ul>
&lt;li>
&lt;p>&lt;a href="https://www.cs.cornell.edu/~shmat/shmat_oak18.pdf">Membership Inference Attacks Against Machine Learning Models&lt;/a>
&lt;/p></description></item><item><title>cmu square</title><link>https://subygan.com/tech/privacy/cmu_square/</link><pubDate>Thu, 18 Jan 2024 00:00:00 +0000</pubDate><guid>https://subygan.com/tech/privacy/cmu_square/</guid><description>&lt;p>SQUARE (Software Quality Requirements Engineering) is SEI&amp;rsquo;s framework for &amp;ldquo;Improving identification, analysis, specification and management (of software)&amp;rdquo;.&lt;/p>
&lt;p>This, Broadly involves 9 steps,&lt;/p></description></item><item><title>data privacy</title><link>https://subygan.com/books/data_privacy/</link><pubDate>Tue, 16 Jan 2024 00:00:00 +0000</pubDate><guid>https://subygan.com/books/data_privacy/</guid><description>&lt;p>These are my notes from Nishant&amp;rsquo;s excellent book, &lt;a href="https://www.oreilly.com/library/view/data-privacy/9781617298998/">Data Privacy: Runbook for Engineers&lt;/a>
&lt;/p>
&lt;h2 id="1-privacy-engineering">1 privacy engineering&lt;/h2>
&lt;ul>
&lt;li>Data Security is usually tightly coupled with privacy. Data security is a base condition to guarantee privacy.&lt;/li>
&lt;li>Data privacy is about how data of users is being accessed and processed.&lt;/li>
&lt;li>There are 4 broad process for data privacy that the author describes,
&lt;ul>
&lt;li>Data classification&lt;/li>
&lt;li>Data inventory&lt;/li>
&lt;li>Data deletion&lt;/li>
&lt;li>Data obfuscation&lt;/li>
&lt;/ul>
&lt;/li>
&lt;/ul>
&lt;h3 id="how-data-flows">how data flows.&lt;/h3>
&lt;ul>
&lt;li>From a Tech leader&amp;rsquo;s standpoint, they need a catalog of data to figure out the tooling to deploy for each case.&lt;/li>
&lt;li>Without the catalog the leader would be flying blind.&lt;/li>
&lt;/ul>
&lt;h3 id="why-privacy">why privacy&lt;/h3>
&lt;ul>
&lt;li>Direct motivators for companies
&lt;ul>
&lt;li>Fines are getting bigger and numerous each year.&lt;/li>
&lt;li>non-investment in the Early stage cases late-stage headaches
&lt;ul>
&lt;li>I liked the Gamebuster example, where IP addresses where being queried by automated scripts and copied in other tables, A clear failure of access control. which led to multiple impacts across the product. My question was, how would a leader be aware of this structure? is getting their hands dirty the only way?&lt;/li>
&lt;/ul>
&lt;/li>
&lt;li>Investigations are usually long and take a lot of the company&amp;rsquo;s and it&amp;rsquo;s leader&amp;rsquo;s valuable time. eg. Bill Gates and his Distraction.
[personal note,] it was interesting to me that all the Data protection examples are from EU and none from America. Show&amp;rsquo;s the fragmented regulation in the US?&lt;/li>
&lt;/ul>
&lt;/li>
&lt;li>There is a lag between engineering and the law, which means engineering has to be prescient about the future laws and it&amp;rsquo;s consequences.&lt;/li>
&lt;li>Regulatory certifications, makes it easier to,
&lt;ul>
&lt;li>Have structure and systems&lt;/li>
&lt;li>Increase trust in the product&lt;/li>
&lt;li>Makes work more efficient, because there are already answers to questions&lt;/li>
&lt;/ul>
&lt;/li>
&lt;/ul>
&lt;p>Early data tagging makes it easier to maintain privacy in the system. If we continue Facebook&amp;rsquo;s &lt;a href="https://www.vice.com/en/article/akvmke/facebook-doesnt-know-what-it-does-with-your-data-or-where-it-goes#:~:text=%E2%80%9CWe%E2%80%99ve%20built%20systems%20with%20open%20borders">ink metaphor&lt;/a>
, With ink that is tagged, you could at least make the ink immiscible and trace it and separate it back. So early tagging is very important. because tags can then be filtered.&lt;/p></description></item><item><title>ai governance and threat modelling</title><link>https://subygan.com/tech/privacy/ml_governance/</link><pubDate>Thu, 09 Nov 2023 00:00:00 +0000</pubDate><guid>https://subygan.com/tech/privacy/ml_governance/</guid><description>&lt;h2 id="how-are-ml-systems-measured">how are ml systems measured.&lt;/h2>
&lt;p>While training this is the loss and accuracy of the prediction.&lt;/p>
&lt;p>For LLMs there are benchmarks like,&lt;/p></description></item><item><title>why privacy?</title><link>https://subygan.com/tech/privacy/why_privacy/</link><pubDate>Tue, 03 Oct 2023 00:00:00 +0000</pubDate><guid>https://subygan.com/tech/privacy/why_privacy/</guid><description>&lt;p>Fundamental problem has been that, privacy is hard to define.
Privacy stretches and shrinks based on context, culture and time.&lt;/p></description></item><item><title>browser privacy test</title><link>https://subygan.com/tech/privacy/browser_privacy/</link><pubDate>Tue, 19 Sep 2023 00:00:00 +0000</pubDate><guid>https://subygan.com/tech/privacy/browser_privacy/</guid><description>&lt;p>I came across this website which tracks and tests privacy of various browsers. &lt;a href="https://privacytests.org/">https://privacytests.org/&lt;/a>

Takeaway IMO:&lt;/p>
&lt;p>in desktop browsers&lt;/p>
&lt;ul>
&lt;li>
&lt;p>Brave, Mullvad and librewolf seems better because of the built in ad protection. but brave clearly has better User experience.&lt;/p></description></item><item><title>privacy articles</title><link>https://subygan.com/tech/privacy/articles/</link><pubDate>Tue, 19 Sep 2023 00:00:00 +0000</pubDate><guid>https://subygan.com/tech/privacy/articles/</guid><description>&lt;h2 id="blogs">blogs&lt;/h2>
&lt;ul>
&lt;li>&lt;a href="https://blog.ovalerio.net/archives/2615">Privacy best practices&lt;/a>
&lt;/li>
&lt;li>&lt;a href="https://www.scu.edu/ethics/focus-areas/internet-ethics/resources/why-we-care-about-privacy/">Why we care about privacy&lt;/a>
&lt;/li>
&lt;li>&lt;a href="https://www.ietf.org/archive/id/draft-farrell-tenyearsafter-00.html">reflections on snowden revealations&lt;/a>
&lt;/li>
&lt;li>&lt;a href="https://martinfowler.com/articles/intro-pet.html">PETs an intro to technologists&lt;/a>
&lt;/li>
&lt;li>&lt;a href="https://www.theprivacywhisperer.com/p/privacy-ux-and-privacy-culture">Privacy UX and privacy culture&lt;/a>
&lt;/li>
&lt;li>&lt;a href="https://www.forbes.com/sites/forbestechcouncil/2022/02/22/whos-good-at-data-privacy/">Who&amp;rsquo;s good at data privacy&lt;/a>
&lt;/li>
&lt;li>&lt;a href="https://cacm.acm.org/magazines/2023/3/270204-metrics-for-success/abstract">Metrics for success: How to evaluate Privacy choice in usability&lt;/a>
&lt;/li>
&lt;li>&lt;a href="https://pair.withgoogle.com/explorables/federated-learning/">How Federated Learning Protects privacy&lt;/a>
&lt;/li>
&lt;li>&lt;a href="https://blog.research.google/2022/02/federated-learning-with-formal.html">Federated learning with formal privacy guarantees&lt;/a>
&lt;/li>
&lt;li>&lt;a href="https://ico.org.uk/media/about-the-ico/consultations/4021464/chapter-5-anonymisation-pets.pdf">PETs&lt;/a>
&lt;/li>
&lt;li>&lt;a href="https://www.scss.tcd.ie/doug.leith/apple_google.pdf">Mobile Handset Privacy&lt;/a>
&lt;/li>
&lt;li>&lt;a href="https://matrix.org/blog/2022/03/25/interoperability-without-sacrificing-privacy-matrix-and-the-dma/">Interoperability without sacrificing privacy: Matrix and the DMA&lt;/a>
&lt;/li>
&lt;li>&lt;/li>
&lt;/ul>
&lt;h2 id="tools">tools&lt;/h2>
&lt;ul>
&lt;li>&lt;a href="https://privacytests.org/">Browser Privacy Test&lt;/a>
&lt;/li>
&lt;/ul></description></item><item><title>a graduate course in applied cryptography</title><link>https://subygan.com/books/applied_crypto/</link><pubDate>Sun, 04 Jun 2023 00:00:00 +0000</pubDate><guid>https://subygan.com/books/applied_crypto/</guid><description>&lt;h2 id="encryption">encryption&lt;/h2>
&lt;p>Shannon cipher assumes a shared secret key that is no smaller than the message being transmitted.&lt;/p>
&lt;ul>
&lt;li>&lt;strong>One time pad&lt;/strong> a Shannon cipher where the keys, messages and ciphertexts are bit strings of the same length&lt;/li>
&lt;/ul>
&lt;h3 id="resources">resources&lt;/h3>
&lt;ul>
&lt;li>&lt;a href="https://crypto.stanford.edu/~dabo/cryptobook/BonehShoup_0_4.pdf">https://crypto.stanford.edu/~dabo/cryptobook/BonehShoup_0_4.pdf&lt;/a>
&lt;/li>
&lt;/ul></description></item><item><title>cicumventing ad block</title><link>https://subygan.com/tech/privacy/gtm/</link><pubDate>Sat, 22 Apr 2023 00:00:00 +0000</pubDate><guid>https://subygan.com/tech/privacy/gtm/</guid><description>&lt;p>using a DNS aliasing to circumvent ad blocking tools got me wondering about the feasibility, because theoretically it should be possible.&lt;/p></description></item></channel></rss>