HNHacker News
TopNewBestAskShowJobs

anuragbiyani

245 karma · joined January 3, 2013

submissionscomments
anuragbiyani··on Show HN: Download all threads of an Orkut community
Can you test the following keystrokes manually and let me know the outcome with your browser:

1) Open any page in chrome

2) Press Ctrl+S

3) Press "Right" arrow key exactly once

4) Type some string, say "-suffix"

5) Press "Home" key exactly once

6) Type in some directory name with "/" at the end, say: "/tmp/"

7) Press Enter.

Sorry if this sounds pedantic, but this is the actual sequence of key stroke the script will mimic in this case, so just want to make sure that work in the first place. Alternatively can you try https://github.com/abiyani/automate-save-page-as script in isolation (with --suffix command line flag).

PS: Under usual circumstances, the name field will should be replaced by something like this by the script (before pressing enter in the "Save as" dialog box): "/fully/expanded/dest/directory/path/<original_page_title>-000X.html"

anuragbiyani··on Show HN: Download all threads of an Orkut community
To save a single page - yes (to be precise, it simulates key presses and does not rely on any mouse action). The saving of the web page part is actually handled by another script I wrote few days back (https://github.com/abiyani/automate-save-page-as), and it was the only option I could find to automate saving a page (as rendered by the chrome). I will be very happy to learn of any other alternative for this task.

This particular script (orkut-community-downloader) uses "automate-save-page-as" script to download every single page, and therefore contain the (simple) logic for navigating through all the posts, polls, and later on changing the links to point to local copy, symlinking stuff to save space, and providing support for "resuming" the download operation.

Do you mind sharing what bugs you are facing (and which platform you are testing on), so I can try and fix them ? If HN comments doesn't seem a fit place, then feel free to email me (ID: <username>@gmail.com).

anuragbiyani··on Timing attack, 6.66% faster
I agree that actual "signal" can still be learned over time, but if we chose additional random wait time from a distribution unknown to attacker (or even better, randomly select the distribution itself, and shuffle often), we can reduce the efficiency of attack dramatically.

But another possible way is to add a deterministic amount of additional wait time (instead of random), i.e., "lie" about computation time of all inputs consistently. If designed properly, this won't leak any signal (assuming the strategy on how to "lie" is chosen randomly itself and shuffled often, so attacker doesn't learn it).

Strategy of waiting for maximum time across all possible requests (as mentioned in your post) is sort of a special case of this, where one (deterministically) add a wait time of W - C (where W = Worst case time across all inputs, and C = actual time taken for this input). But we can use a less aggressive approach to hide the actual computation time as well.

anuragbiyani··on Timing attack, 6.66% faster
A server can easily add an extra random wait to all unauthorized responses, making such attacks almost infeasible (at least for web-app case).
anuragbiyani··on World Cup results for hackers
Crossposting from: https://news.ycombinator.com/item?id=7908081.

A small bash script to display the current match scores as part of the terminal prompt: https://gist.github.com/abiyani/34f25dc8ed2a862517e7

Sample output:

  abiyani@vaio:~ [BEL-2-1-ALG, RUS-1-1-KOR, BRA-0-0-MEX]
anuragbiyani··on World Cup data in JSON
Created a small bash script to display the current match scores as part of the terminal prompt : https://gist.github.com/abiyani/34f25dc8ed2a862517e7

Sample output: abiyani@vaio:~ [BEL-2-1-ALG, RUS-1-1-KOR, BRA-0-0-MEX]

← PreviousPage 2 of 2