The initial response was very quick, but honestly: Fixing such a simple XSS hole on a web application shouldn't take 7 days. In the age of automated testing and continuos deployment, escaping that parameter is trivial.
Granted. If you have a systemic issue and you need to fix your framework to always escape unless told otherwise, that will take longer, but it won't prevent you from quick-fixing just the reported issue.
I had to do a double-take when I read the "7 days" part, because I wasn't sure why that should be a pleasant surprise.
If it were their desktop client, that kind of turnaround would be understandable. For their website? I don't see why they couldn't have deployed a hotfix immediately.
On the other hand, perhaps they had a quota of more serious security issues and were preoccupied with addressing those.
It's likely that it was fixed sooner than that. I would bet they gave it 7 days before he could talk about it, so they could audit the rest of their site.
"We have fixed it" is easy to understand as "you're now free to blog about this flaw"; one could imagine why Spotify would want to take the time to properly audit the rest of the site.
Also, Spotify is a serious business that doesn't want to go down, and likely has a testing/deployment process a bit more complicated than "just push to master".
Indeed. I would like to add that the initial response was most likely automated (looks kinda generic). Kinda makes me want to send a dummy email and see if I get a similar response as fast.
But in all fairness, 7 days is still acceptable. It's far from "best ever", but compared to others it's still great.
Yes, Spotify is a Swedish companies, but we have several engineering offices in the US and in Europe. The current openings are listed on spotify.com/jobs
From OP:
"Spotify came to my attention because of that specific page as I'm currently looking to build a security portfolio."
This is almost as foolish as it is illegal.
Finding web security issues is much more complicated legally than finding security issues in desktop/server software. I can download and install Apache and test it on my local box and discover things like the Range Header DoS. There are no legal issues for me to worry about.
I can't download and install "LinkedIn" and test it for web issues. I have to test against their site running on their computers. Immediately in the US the Computer Fraud and Abuse Act comes into play (though the OP lives in Denmark). "Hmmm, let me 'test' random website X for something that could cause a DoS" is insane.
Even if a company has a formal "we are OK and authorize people to test our site for security issues" policy that is still extremely dangerous from a legal perspective. Maybe you are only authorized if you follow their specific disclosure policies. Maybe you are only authorized to do non-destructive testing. Every company is going to have some tipping point where, if you testing impacts X or caused Y downtime, that's enough damage where they will seek your prosecution.
Building a "security portfolio" against websites is a stupid idea with some potentially huge negative consequences. If anyone insists on doing this, at least go looking for issues in projects that you can download and audit locally.
Spotify agrees on security researching as long as it's not disrupting the service. Any disrupting analysis method is forbidden and I would not proceed in auditing something without permission before having a security audit contract signed by the company operating the service.
"Hmmm, let me 'test' random website X for something that could cause a DoS"
The author was testing reflected XSS which is inherently client side, there should be no case of concern for DoS here.
"Building a "security portfolio" against websites is a stupid idea with some potentially huge negative consequences."
The author mentions Apple, Linkedin, Amazon and AT&T all of which have some type of vulnerability notification program. (See https://bugcrowd.com/list-of-bug-bounty-programs/ for a great list of programs) I would have to say that as long as the researcher was performing within the scope of the respective program, there should be no worry. I think it's the exact opposite of a stupid idea. Building out a portfolio within the scopes of the programs is a great way to build some security reputation.
I understand what you are saying, but I disagree with the risk/reward.
Yes, the author was testing with a reflected XSS attack string, and yes "there should be no case of concern for DoS". But that's exactly my point. You have no idea what effect that will have.
I personally have caused an inadvertent DoS attack with an XSS probe while doing a security audit for a large SaaS company. (The attack made the code throw and exception and some of my escaping characters caused havoc with their error logger, sending it into a infinite loop trying to log data, pegging the CPU and growing memory until the disk thrashed.)
You don't know what you actions will cause because the server/app is a black box. And if you want to audit that, you should have a formal agreement with the company that they know what you are doing and its OK. A generic, flimsy, non-personal "everyone can try and 'hack' us and it's OK" policy published somewhere is just too little protection, especially given how the CFAA is applied in courts today.
I think bug bounties can make the risk/reward ratio palatable, by either lowering risk or increasing the reward.
If a company offers monetary awards for web vulnerabilities in their site, that's a good sign. It shows they recognize the monetary value on what you are doing and made the decision to reward you. It shows the extent of their commitment to the bounty program that they have allocated budget and have accepted the added accounting challenges to provide money for issues. It also means that have thought through exactly what they are expecting and what they are willing to pay.
In my eyes, someone offering marketing swag or "putting your name on a web page" doesn't think your work is as valuable. Thus they are more likely to be angry and seek legal remedies should your testing cause them harm in some way, because in their eyes the value of your security testing or your findings doesn't out weight the damage.
In short, all of this is an indicator of how much thought was put into the company's policy on finding and reporting security issues, and how mature their reaction will be to things they have not thought of or bad things happen. The less mature, the more likely they are to decide your testing was "unauthorized" and to seek legal remedies.
Compare Spotify's policy [1] with Google's Policy[2]. Which one has put more thought and planning into running a program like this? Obviously people's risk/reward appetites differ...
Please note that I am taking the risks of bug bounties/"its ok to audit us" policies that apply public websites. The Stripe/Matasano challenges are specially created systems and are always a great thing to do.
>The attack made the code throw and exception and some of my escaping characters caused havoc with their error logger
Heh, something similar happened to me during a recent audit. I didn't even know until an admin emailed me saying that I'd broken a bunch of batched jobs, and not to test that until it was fixed.
>A generic, flimsy, non-personal "everyone can try and 'hack' us and it's OK" policy published somewhere is just too little protection
It hasn't worked out too badly for me. I stay away from industries with lots of "suits" (banking, etc,) but if a company is implicitly encouraging independent pentesting by publicly crediting reporters, and you act in good faith, I can't see any charges sticking.
Totally agree with that. The important time is the one between the disclosure and the fix, not the one between the disclosure and the first response. It could be a bot :)
We work hard on becoming better and faster at tackling these and other issues. We try to learn from incidents and mistakes in order to be faster next time.
(You probably already saw this but if you didn't)Spotify employee already pointed out the bug was fixed within the day the original email was sent. 7 days was probably after an audit to check for the issue in other places and then after 7 days they felt comfortable letting the researcher talk about it. Cheers!
24 comments
[ 2.7 ms ] story [ 67.6 ms ] threadGranted. If you have a systemic issue and you need to fix your framework to always escape unless told otherwise, that will take longer, but it won't prevent you from quick-fixing just the reported issue.
If it were their desktop client, that kind of turnaround would be understandable. For their website? I don't see why they couldn't have deployed a hotfix immediately.
On the other hand, perhaps they had a quota of more serious security issues and were preoccupied with addressing those.
Also, Spotify is a serious business that doesn't want to go down, and likely has a testing/deployment process a bit more complicated than "just push to master".
But in all fairness, 7 days is still acceptable. It's far from "best ever", but compared to others it's still great.
Regarding the 7 days patch, that does look long. Not sure why. It'd be interesting to find out.
By the way, I like your resume. Visually impressive; a pleasant to read. It'd be interesting to see that being a shirt.
This is almost as foolish as it is illegal.
Finding web security issues is much more complicated legally than finding security issues in desktop/server software. I can download and install Apache and test it on my local box and discover things like the Range Header DoS. There are no legal issues for me to worry about.
I can't download and install "LinkedIn" and test it for web issues. I have to test against their site running on their computers. Immediately in the US the Computer Fraud and Abuse Act comes into play (though the OP lives in Denmark). "Hmmm, let me 'test' random website X for something that could cause a DoS" is insane.
Even if a company has a formal "we are OK and authorize people to test our site for security issues" policy that is still extremely dangerous from a legal perspective. Maybe you are only authorized if you follow their specific disclosure policies. Maybe you are only authorized to do non-destructive testing. Every company is going to have some tipping point where, if you testing impacts X or caused Y downtime, that's enough damage where they will seek your prosecution.
Building a "security portfolio" against websites is a stupid idea with some potentially huge negative consequences. If anyone insists on doing this, at least go looking for issues in projects that you can download and audit locally.
You are reaching a bit too far, it's not like he hacked their database or accessed private data.
"Hmmm, let me 'test' random website X for something that could cause a DoS"
The author was testing reflected XSS which is inherently client side, there should be no case of concern for DoS here.
"Building a "security portfolio" against websites is a stupid idea with some potentially huge negative consequences."
The author mentions Apple, Linkedin, Amazon and AT&T all of which have some type of vulnerability notification program. (See https://bugcrowd.com/list-of-bug-bounty-programs/ for a great list of programs) I would have to say that as long as the researcher was performing within the scope of the respective program, there should be no worry. I think it's the exact opposite of a stupid idea. Building out a portfolio within the scopes of the programs is a great way to build some security reputation.
Yes, the author was testing with a reflected XSS attack string, and yes "there should be no case of concern for DoS". But that's exactly my point. You have no idea what effect that will have.
I personally have caused an inadvertent DoS attack with an XSS probe while doing a security audit for a large SaaS company. (The attack made the code throw and exception and some of my escaping characters caused havoc with their error logger, sending it into a infinite loop trying to log data, pegging the CPU and growing memory until the disk thrashed.)
You don't know what you actions will cause because the server/app is a black box. And if you want to audit that, you should have a formal agreement with the company that they know what you are doing and its OK. A generic, flimsy, non-personal "everyone can try and 'hack' us and it's OK" policy published somewhere is just too little protection, especially given how the CFAA is applied in courts today.
If a company offers monetary awards for web vulnerabilities in their site, that's a good sign. It shows they recognize the monetary value on what you are doing and made the decision to reward you. It shows the extent of their commitment to the bounty program that they have allocated budget and have accepted the added accounting challenges to provide money for issues. It also means that have thought through exactly what they are expecting and what they are willing to pay.
In my eyes, someone offering marketing swag or "putting your name on a web page" doesn't think your work is as valuable. Thus they are more likely to be angry and seek legal remedies should your testing cause them harm in some way, because in their eyes the value of your security testing or your findings doesn't out weight the damage.
In short, all of this is an indicator of how much thought was put into the company's policy on finding and reporting security issues, and how mature their reaction will be to things they have not thought of or bad things happen. The less mature, the more likely they are to decide your testing was "unauthorized" and to seek legal remedies.
Compare Spotify's policy [1] with Google's Policy[2]. Which one has put more thought and planning into running a program like this? Obviously people's risk/reward appetites differ...
Please note that I am taking the risks of bug bounties/"its ok to audit us" policies that apply public websites. The Stripe/Matasano challenges are specially created systems and are always a great thing to do.
[1]- https://www.spotify.com/dk/about-us/contact/report-security-... [2]- http://www.google.com/about/appsecurity/reward-program/
Heh, something similar happened to me during a recent audit. I didn't even know until an admin emailed me saying that I'd broken a bunch of batched jobs, and not to test that until it was fixed.
>A generic, flimsy, non-personal "everyone can try and 'hack' us and it's OK" policy published somewhere is just too little protection
It hasn't worked out too badly for me. I stay away from industries with lots of "suits" (banking, etc,) but if a company is implicitly encouraging independent pentesting by publicly crediting reporters, and you act in good faith, I can't see any charges sticking.
We work hard on becoming better and faster at tackling these and other issues. We try to learn from incidents and mistakes in order to be faster next time.
Thanks for the feedback.