Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Apply HN: Vaultedge – a private Google for your private data

news.ycombinator.com

Post reply on HN
Print view View original

1–10 of 28 posts

Apply HN: Vaultedge – a private Google for your private data

#1

Post by sajeevaravind »

Vaultedge demo : https://www.youtube.com/watch?v=ufnMZjg5344

Problem : Consumers have data in many different places like cloud storage, emails, laptops and offline devices. Trying to find a document which we created or downloaded few months or even days ago will require answering questions like was I at home or office while downloading it, did I receive it over an email or did I store in dropbox? You quickly make an assumption and search for the filename or keywords there and if you don't find it you search in other places until you find it or you give up.

Vaultedge : is an attempt to help find your document quickly just in few clicks. I personally faced this problem in my previous job as Software Architect and asked people around on what tools they use. Leaving aside few who were well organized, rest of the folks faced the same issue. They would love to have a tool which will allow search across their online and offline content contextually like to search for "last three months of cell phone bills" or "the document on networking send by x". They were also interested in a grouping capability which groups documents into similar categories so that they can see - all their bills in one place - to highlight the difference between the fluctuating cable or cell phone bills month over month - to download bills for a month in bulk - easily find all tax records for a year etc

With those two functionalities (Search & Grouping) in mind we are building Vaultedge. Contextual search is something we are still building.

About us: I have spent all my life (~16 years) building enterprise products for Storage, Data and Analytics, my last job being Architect for Backup products at NetApp. I'm a sole founder and backend developer now, but in talks with another tech guy to join as co-founder. I also have a awesome frontend engineer in my team as employee.

Love to answer your questions, comments and to take feedback.

Re: Apply HN: Vaultedge – a private Google for your private data

#2

Post by brudgers »

How do you anticipate users inputing the metadata necessary to make this work?

For example, how will the service know how to access my phone bill /\ my taxes /\ the brown cat playing Chopin on the piano video?

Not so much interested technical details as to the degree of effort an ordinary person with an iPhone would need to invest to make use of the service.

Good luck.

Re: Apply HN: Vaultedge – a private Google for your private data

#3

Post by sajeevaravind »

brudgers wrote: post #2

How do you anticipate users inputing the metadata necessary to make this work? For example, how will the service know how to access my phone bill /\ my taxes /\ the brown cat playing Chopin on the piano video? Not so much interested technical details as to the degree of effort an ordinary person with an iPhone would need to invest to make use of the service. Good luck.

User has to connect his services to Vaultedge. Services can be cloud storage like dropbox and google drive, emails like gmail and outlook and offline storage like laptops and usb disks. Vaultedge will periodically check for any new files (docs only now) in these services and classifies them using Machine learning and also builds a search index for these files. Other than giving oAuth read access to these services, user need not do anything.

Thanks for asking. Happy to answer any follow up question you may have.

Re: Apply HN: Vaultedge – a private Google for your private data

#5

Post by brudgers »

sajeevaravind wrote: post #3
brudgers wrote: post #2

How do you anticipate users inputing the metadata necessary to make this work? For example, how will the service know how to access my phone bill /\ my taxes /\ the brown cat playing Chopin on the piano video? Not so much interested technical details as to the degree of effort an ordinary person with an iPhone would need to invest to make use of the service. Good luck.

User has to connect his services to Vaultedge. Services can be cloud storage like dropbox and google drive, emails like gmail and outlook and offline storage like laptops and usb disks. Vaultedge will periodically check for any new files (docs only now) in these services and classifies them using Machine learning and also builds a search index for these files. Other than giving oAuth read access to these services, us…

Sorry for not being clear, I am curious regarding how my phone bill becomes searchable.

Re: Apply HN: Vaultedge – a private Google for your private data

#6

Post by sajeevaravind »

brudgers wrote: post #5
sajeevaravind wrote: post #3

Earlier quoted context omitted.

User has to connect his services to Vaultedge. Services can be cloud storage like dropbox and google drive, emails like gmail and outlook and offline storage like laptops and usb disks. Vaultedge will periodically check for any new files (docs only now) in these services and classifies them using Machine learning and also builds a search index for these files. Other than giving oAuth read access to these services, us…

Sorry for not being clear, I am curious regarding how my phone bill becomes searchable.

Short answer is Vaultedge will create search index based on the content of the bill and also has additional knowledge that it is a cell phone bill from its classification algorithm. Put together this will allow user to search for "last month's cell phone bill".

Long answer: Let's say your bill is in pdf or html format. For Vaultedge initially it is nothing but a document, it has no idea that it is a bill. Vaultedge will extract the contents of this document and test it against our training data set using Machine Learning(ML). In this process, the document will be identified as a "Bill". Then it will apply it on a different training set which will identify it as a "cell phone" or "cable bill". Then we do further analysis and extract info like this bill is "for the month of march 2016". Now all of these info is used to construct the search index. So you can search now for "march bill" or "last months bill" to get to that bill. Please don't hesitate to clarify if anything is not clear.

Re: Apply HN: Vaultedge – a private Google for your private data

#7

Post by sajeevaravind »

vit05 wrote: post #4

I think it would be very risky to put so much information in a single place that can sift through all these details and summarize this information about you. What are the security measures that you will offer?

Vaultedge will not store the actual files, only metadata about the files. We plan to encrypt the metadata and also optionally allow user to control the keys. We will also have 2-factor authentication. Please suggest if you have any suggestions. Thanks for asking.

Re: Apply HN: Vaultedge – a private Google for your private data

#8

Post by vit05 »

sajeevaravind wrote: post #7
vit05 wrote: post #4

I think it would be very risky to put so much information in a single place that can sift through all these details and summarize this information about you. What are the security measures that you will offer?

Vaultedge will not store the actual files, only metadata about the files. We plan to encrypt the metadata and also optionally allow user to control the keys. We will also have 2-factor authentication. Please suggest if you have any suggestions. Thanks for asking.

this is better. My suggestion is that the metadata is not linked to a complete profile.

Re: Apply HN: Vaultedge – a private Google for your private data

#9

Post by sajeevaravind »

vit05 wrote: post #8
sajeevaravind wrote: post #7

Earlier quoted context omitted.

Vaultedge will not store the actual files, only metadata about the files. We plan to encrypt the metadata and also optionally allow user to control the keys. We will also have 2-factor authentication. Please suggest if you have any suggestions. Thanks for asking.

this is better. My suggestion is that the metadata is not linked to a complete profile.

I would like to understand a little better your comment about "metadata is not linked to a profile". Do you say by not linked to a profile, there shouldn't be a way for a hacker to figure out the metadata of a user? Or you mean "metadata" should be associated to individual accounts(say dropbox), so incase of a breach, only the metadata for that account is compromised and not for all the accounts of a user? Thanks for your time.
Post reply on HN