Shifting Left and Right – Rickard Carlsson, Detectify
The shift left movement is nothing new. DevSecOps have been conducting security tests earlier in the development process for years – but have headlines of successful attacks slowed down? The problem is this – the notion of shifting left is dependent on a standard linear development process. In reality, development is anything but linear. Effective DevSecOps requires testing in both staging and production environments, particularly with attackers increasingly targeting unknown, forgotten, and neglected assets. There are advantages to implementing both a shift left and right approach. Testing in development is important, but isn’t a silver bullet and it’s also critical to test in production. Shifting everything left, and acting like that fixes everything, is a mistake.
Transcript
This is Textron TV. Hey everyone, welcome back to techstrong TV. Our next guest is a new First Time guest on on our show.
And I think it's the first time we featured his company here. I'm really excited to talk to him. His name is Ricard Carlson and Ricard is the CEO of a company called detectify.
If I got that wrong record, please correct me. But if not, welcome. Thank you.
Nice to be here and checking nice. Nice to have you so Ricard before we jump into what we really want to dive into today. It is the first time on for our audience our first exposure.
Why don't you if you don't mind. Can you give us a little company background on detectify maybe a little bit of your own background? of course happy to do so and I guess my background comes I can the company background is my my technical co-founders started off as Really strong and capable white hats so they were hacking companies under responsible disclosure and Bug Bounty programs and they started to realize that quite a few years ago that there most of security regime was done and focusing very much in infrastructure security back.
Then I mean almost seven eight years ago. I'm very little Focus respect and spent on application security and they started automating a scripting their own attacks. In ways to simplify their own bug bounty hunting and eventually that turned into a product or starting to get into a product and my background.
I'm a physicist software engineer mathematician by from from school University went to management consulting but then realized I actually want to get back to technology. I think that with them just so to say run everything related to product and Commercial side in the beginning and then they continue building on the engines. excellent That's actually a great background, you know in surprisingly insecurity.
It's not. Unusual right to have a group of two three founders where one or two of them really are very technical folks. You have business and operations expertise.
Sometimes you have sales right marketing as well. How long is it been in business website stuff like that? Yeah, we've been around for about eight years, but taking a bit of shift throughout these years.
I guess we started off the living building and very much of a traditional web app security scanner, but that was built for modern companies that wanted to leave. So say the infrastructure providers that they did in the past. And but fairly early on we started also working on the side and different types of attacks related to dns-related vulnerabilities.
We wrote I think the first one of the first blog posts out there on attack methods related to misconfigurations related to subdomain takeover already back in 2014. We built the first prototype of this service testing for some of the main takeovers at scale. And 2016, you know, we wanted to customers that we built it and then gradually we realized we want to leave so to say the old traditional web apps that game behind us and fully transitioned into what we're working on now is related to external attack surface management.
And this yeah, this comes back since since we started in 2016 that this is what we wanted to do, but it's only to feel like the market starts to now considering their attack surfaces as a way to is a way to unhealth should do security with the combined with modern development practices. You know, it's interesting. It's actually the second interview I've done today.
That is talking about attack surface management attack surface security. And part of it is is that you know, our tax service is today are so different between covid the cloud iot and all of these things our attack surfaces. Are it's just not the people's laptops and the servers anymore.
You know what I mean that looking back on it that seemed almost simple right? Just super simple. When you had when you had your service in the basement.
You had one ethernet cable going out through your firewall and you to a hundred percent knew what you're exposing and For someone to open a port in that fire will in the basement you have to fill out the form and it took you one month. Today when in the devops world when people are controlling their whole exposure through terraform and with you know, the DNS pointers. No one that the rate of change of the tax surface of companies are so vastly different compared to where they were, you know a few years ago.
Absolutely, and the old adage of you can't protect what you don't know you have. right, and so the whole kind of just the A subset of this is just a tax surface Discovery. I said Discovery knowing what you know, what what you're exposure is before you can even think about protecting.
It is another just a whole world going to itself. And I think the and especially they come we see many of our most advanced. customers and the way that we think about is that they They don't want to hunt for vulnerabilities.
They want to hunt for also, like policy anomalies and violations of their own security regimes. As an example. We have a customer that says okay, we have 200 domains and all of them should be behind cloudflare and we I mean as our example that we can run a test for them as a special test that say, okay, we can validate that policy every eight hour that all of these 200 domains are behind cloudflare and this is not something that you can really readily do from the inside or if you're thinking from a static code analysis tool or something.
This is about how you set up your configuration and what you are Tax Service looks like Absolutely. I don't mean to cut the conversation short, but I wanted kind of segue into you know, what we want to talk about today because another thing that's happened is look I've been in I was in security 25 years still am in security 20. Congratulations.
Yeah. Thank you. So.
And I started a company called the still secure co-founded in 2001 and we had an early entrance in the vulnerability management space right doing they were it was in that space scanner. Yeah. But back then just getting people to scan their production environment.
Was was quite a task if you scanned once a year man, I loved you right because you were ahead of most of the other people, you know, one of the big shifts no pun intended. We've seen his you know, what the rise of deaths Tech Ops is this so called shift left. Movement right shift left idea where hey, not only are we going to scan in production?
We're actually going to scan and development because kind of a stitch in time saves nine, right? It's much cheaper to fix vulnerabilities pre-production than post and so by China automate and and move move your moving our scanning left shifting it left. We we have we release a higher quality code and as we go to Automation and more frequent code releases, you know automating that and shifting it left.
Is is a good thing? Now of course, well, I'm gonna stop there and say what do you think regard? I I think that it's an it's something that's that's actually I would say that it's almost completely wrong to say that it's cheaper to find things earlier that I think was true when you were coding for three months and then you did a release because then you have built up so much.
Surrounding things around something that actually really dependencies that you need. If you need to revert something was extremely costly if you look at most modern companies or if you take fully fully integrated domain driven engineering teams that have all the capabilities that you know to host code ship release in their own in within their own team. They shift maybe multiple times per day the whole thing about it should be 10x cheaper because you found it earlier.
I think that's wrong because earlier does not exist because you might do an experiment on 2% of the traffic that you receive. That's an example or you say You release something on their feet your flag, and then you're reverted back if you didn't work. So I think the description the difference between stating and production.
There's no longer exist in reality. It only existing compliance documents. Fair, I think another and maybe you're saying it without saying it another.
Quite frankly, you know, we used to have Dev environments test environments where we would test features and code before going in a production. I I think in the rush to to release the deploy today. You know our production environments quite frankly are tested environments or you know to a certain extent you would feature Flags many industries that are not regulated because of deep compliance where I mean some Financial Industries where you're not allowed to run unvalidated code around tested code in production, but for most of the companies, yeah production is the new testing environment.
Yeah. So I think that's one aspect of the issue. The other aspect of the of an issue is You know historically that's in the security world.
We tend to be focused on whatever the shiny new object is and and in in putting our focus on shift left. We we almost ignore or or discount or you know, we just don't give the emphasis on so-called shift, right right doing that post-production type of testing that we would just getting good at what we started. I don't getting good at some relative term.
Well, we would doing better at then. We before we you know started with the shift left as well. You see that as a problem.
I see it. There's a problem that people or organization are trying to hunt for serial vulnerabilities and production. Because if you're gonna achieve that you will be implications on your overall business in terms of speed and other business.
I think it's it's too high but I think what organization should really focus on is if a critical thing is being found in production because also but I think it's typically said 80% of the code that you exposing is not being actively maintained. So if you only scan if you're production, how do you then make it get access to it? But in general if you find you problems in vulnerabilities in production how fast will you be able to review it and ship on you and you attached to it?
I think the best thing class organizations then we actually talking of ours when organizations that we work. We have some organization, you know, that's sitting there and they they have a review cycle. They have security boards and they everything need to go through different Gates and it takes them three four weeks.
Maybe if they are faster resolve something I think customers that says if you do it in 90 days, we're good, but if instead of getting and Agile enough process if something is found in production that you can fix the amount of hours. Then I think you are more better Geared for when new things happen. Then you know when for example, yeah a low for J or spring for Shell or the new thing so say s*** hits the fan you will be have the systems and processes ready to actually act on things fast.
Absolutely. So record, you know, you said something about the zero. Zero vulnerabilities even ship software as a goal.
You know what and if I said that to a hundred people probably 90 of them would say, yeah, that's a great goal. I mean who wants to ship with vulnerabilities? but to me, that's almost like And I don't mean it in a light way or I'm not being disrespectful but it's almost like China's zero covid policy, right?
You're not going to just totally stop covid. From a billion and a half people over there. So to kind of just like shut down everything every time one person gets covid or whatever is Is after a while I think it shows that it's not the right way to do it.
And I think it's the same thing with the zero vulnerability in shipping code, right? Security is always been. If we want the great security just unplug everything from the internet.
You're okay, you know, you'll be pretty secure but Security's always been about managing risk. Yeah, not necessarily especially from a business point of view how much am I gonna spend versus what's my risk reduction or exposure and You know, so I would say can you get to a zero vulnerability shipping? I don't know if you could ever get there, but the price of getting close like getting to the speed of light right the closer you get to that the more expensive the more energy.
It takes the more everything it takes. when you do a risk You know rewards kind of thing. That's impractical.
Yeah anyway, because they should do that. It's crazy. I think but most people would agree.
I mean if you 99 or 90 people would say that it's a great thing. But I typically I would ask typically people three or four questions does that I mean do you really have access to all pipelines and service and Cloud environments? You could test everything in a s*** left mentality and second thing, you know could all security tests or security is just be the tech that when you're testing or doing cold review on an isolated component and how well does your staging environment really represent your production environments and how much delay can you introduce into a cscd pipeline before developers start screaming at you?
So All of these four things I think also says that it's impractical to do it. Well, so I guess this is where you and I are probably going to disagree then though. So I I want Goldilocks.
Right. I I want to find that sweet spot where I do enough shift left testing that I find a good amount of vulnerabilities before before this stuff goes into production, but I want that shift last testing to be fast automated and help improve my quality. I don't want it to take away.
From testing that I do in post-production either because to me that's almost like a different animal. That's a different. You know different set of tests the different focus and everything else.
I want my cake and eat it too. I want both. I agree with you.
I agree with you that we agree. Yeah, because but you should use the right tool for the right job. I think running static code on us is or social sc8 software composition analysis tools as part of like a staging a pipeline makes a ton of sense, you know, they are fast they they run directly the developers get direct feedback, but then we started to talk with organizations that wants to introduce Dynamic testing in staging environments.
And I think that's where people are over engineering and making it way to complicate it because we're if we're because where are they heading? So I agree with you you can have your things in the on the left side and you should because that makes sense, but then you have people that are trying to Move some of that they said they want to run everything on that side. And I think that's where the the cost versus benefit is.
It's not maybe always that clear. You know to go all Star Wars on you. There has to be balance in the force.
We you needed you know on both sides. of Record, this is a great conversation where we're just about at a time for people want to get more information about detective fight. Where do they go?
They take to find outcome. Can you spell it for us? com.
Perfect, man. Hey, thanks for jumping on here on texture on TV with us. Today.
We'd love to have you back on you know, these are the kinds of topics we discussed in a lot of our virtual events and everything. We've got to get you on a panel with some folks and let's talk about this. Thank you.
Thank you. com. Check it out record Carlson CEO over at detectify.
Thanks for joining us. We're going to take a break on Tech strong. We'll be back in a minute.