In January the Conference on Innovative Data Systems Research (CIDR) was held, and I dedicated a week of The Morning Paper to reviewing some of my favourite papers from the programme.
For February, I’ve chosen two conferences to bring you papers from:
- NDSS 2015 – The Network and Distributed System Security Symposium held in San Diego from February 8th-11th, and
FAST’15 – The 13th USENIX Conference on File and Storage Technologies to be held in Santa Clara this week (22-25th Feb.).
There are plenty of good papers to choose from between these two, and in keeping with other themed weeks on The Morning Paper I’m going to reveal my selections ahead of time. Write-ups will be announced daily on twitter as normal (#themorningpaper), but the best way to make sure you never miss an issue is to subscribe to the mailing list.
- Information-Flow Analysis of Android Applications in DroidSafe – Gordon et al. An incredibly effective use of static analysis techniques to detect security leaks from Android applications.
- Machine Learning Classification over Encrypted Data – Bost et al. Showing us that it is possible to use classifiers to make predictions while keeping the client’s data hidden from the server, and the server’s model hidden from the client. Plus it uses homomorphic encryption which is a very impressive technology!
- Bloom Cookies: web search personalization without user tracking – Mor et al. A great application of Bloom filters that shows it is possible to get the benefits of personalized search results without losing privacy.
- F2FS: A new file system for flash storage – Lee et al. We know flash has the potential to be disruptive – but what does it really mean down at the file system / data store level? And if you designed a file system specifically for flash, what kind of performance improvements might we be talking about?
- RIPQ: Advanced Photo Caching on Flash for Facebook – Tang et al. How Facebook combined the performance of flash with new approximate priority queue frameworks that work efficiently on the medium by avoiding large numbers of small random writes, and improved their hit ratios by 20% at the same time!