Hello and welcome back. I hope you're looking forward to another fantastic talk here at the
AppSec Village for DEF CON 29 in safe mode. It's important to remember you're 3-2-1.
Even though we are separate from each other and socially distanced, your cohabitants will really
appreciate that one in there. And just take care of yourselves. Get some good sleep, get some good
food, make sure you take a shower. For our next talk, we've got David Waldrop talking about the
DevOps and Agile Security Toolkit. After 25 years in the application development, David moved to
information security just over seven years ago. His passion is helping developers write secure
code. He's currently an information security advisor supporting the application development
community at his employer. Let's put our hands together and welcome David Waldrop.
Hi and welcome to the AppSec Village. I hope you're having a great DEF CON 2020,
albeit in safe mode. My name is Dave and for the next 30-45 minutes, we're going to talk about the
DevOps and Agile Security Toolkit. Now before I get started, I have to make the standard
disclaimers. First, this presentation and any comments that I make are my own and do not
necessarily represent those of my employer. Secondly, I will mention a few example products.
Some are commercial, some are open source. These are not necessarily meant as recommendations.
They are mentioned to serve as a starting point for your own investigation and review.
These are things that I've run across in my journey, so take that as it is. Let's start
with the two terms that are on everybody's buzzword bingo sheet these days, DevOps and Agile.
We need to get definitions in place for the duration of this talk, so we have an idea that
we're all dealing with the same playing field. Agile. Agile focuses on continuous iterative
development and testing. Software is developed in smaller increments and then is integrated
for final testing and delivery. Scrum, Kanban, and XP Extreme Programming are examples of Agile
approaches. On the other hand, DevOps focuses on the collaboration between app dev and operations.
The focus is on deploying code to production in a fast manner, usually automated.
So you can see from the definitions that these are two separate concepts. However,
many organizations tend to combine them into one initiative. Very, very rarely do you find
a shop doing DevOps and not doing Agile, or vice versa. So no matter whether they start with DevOps
or they start with Agile, the one piece that's usually late to the party is security. But no
matter how late security is invited to the party, or whether we have to crash the party ourselves,
we usually can still bring valuable resources without slowing things down, or at least without
slowing them down a lot. And that's what we're going to talk about through the rest of this
discussion. So for the past few years, I've worked with several different Agile and DevOps teams.
My integration with this started as a pilot program where I was the information security
representative that was assigned to help introduce Agile to our organization. More recently, I've
implemented the information security integration with the Agile and DevOps teams across our
organization. During this session, I'm going to share with you some things that really worked well
and some things that just didn't seem to work at all. So hopefully you can take something
away from this and apply it to your own organization.
This particular slide will give you the map of what we're going to talk about for the rest of
this discussion. We're going to talk about Agile staffing. We're going to talk about the eight
simple questions. We're going to talk about the security champions program, and then go into
developer training, which is my favorite part of my job. Security in the software development life
cycle. And then we're going to hit the hat trick of securing your code. Static code analysis,
dynamic code analysis, and open source analysis. And then finally, we're going to wrap things up
by automating the heck out of it. So let's start with Agile staffing.
Sherman set the way back machine from when I started with Agile. One information security
person was assigned to one Agile team. That security person would be integrated into that team
and attend every ceremony, every meeting, everything, every day. That didn't work.
There are a couple of reasons I think that that didn't work. First of all, Agile teams tend to
multiply. You can go from a proof of concept with one team to having close to 30 teams in less than
two years. Most organizations do not have the information security bandwidth to support that
type of staffing model. In addition, the scrum masters tend to treat every member on the team
as though they can perform all tasks. And when you're dealing with a team of all developers,
that makes perfect sense. This model tends to fall apart though when you put information security,
QA, middleware, any other infrastructure or shared services team as a part of the Agile team.
What happens is in the Agile process, their hours count against your available development hours.
But the infrastructure people, the information security people do not have the skills or the
capabilities to do some of the development. So you kind of see that there's a mismatch. You're
given hours, but you're not allowed to use them. So it caused a lot of problems for some of our
scrum masters early on. Again, we realized this was not the way to do it. So I ended up coming up
with a two-tiered approach for trying to integrate information security into the Agile team.
First, at a high level, at the enterprise level, we will have an architect, an information security
architect, and they will inject into those high-level projects and basically set up guardrails
and guidelines. So as projects are being devised at the visionary level, they can jump in and try
to help guide them into making more secure decisions at that high level. Now when you get
down into the weeds though, each Agile team does need to have somebody they can rely upon
for information security. And so what we ended up doing is we created named resources within
the information security team that are assigned to Agile teams. Now it's not like the previous
model. You're not assigned, you're not sitting with them, you're not assigned to go to every meeting.
We're going to talk about how we scaled that back, but we ended up finding a way where based on
where information security can add the most value, we attend those ceremonies and only those
ceremonies. So at the end of the day, each information security representative can support
between six and eight different Agile teams. That is a model that is certainly sustainable.
So what is a little different? Well first of all, we make sure that every Agile team has a named
resource that is responsible for information security for them. So this gives them a named
contact. If they ever need something, they know exactly who to call and where to start that
quest. Secondly, we attend, or the Agile reps attend, the backlog refinement sessions. These
are meetings that happen once every two weeks before the next sprint, and it defines what
they're going to work on in that next sprint. It's a perfect opportunity to jump in and say,
you're planning this work, we need to think about blah blah blah from an information security
standpoint. We can get in just ahead of when development starts. It's a perfect time to bring
up security issues, make recommendations, and answer questions. And then finally, we attend
the quarterly planning increment meetings where the vision is cast for the next quarter, and we
can make sure that we're on their radar for things that we think we need to be a part of.
So this is kind of a high-level diagram. The purple arrow at the top is the information security
architect. That particular person, as you can see from the diagram, sets guardrails for enterprise
planning projects. All of the red arrows are the different places where our individual security
representatives can or may inject into the Agile teams. We don't inject in all those places,
but there are places where we can, and we give them that option to call us if they're needed.
So let's move on to what I like to call the eight simple questions.
This particular concept came up as a result of assigning our Agile reps to their teams.
We found that the leadership within the Agile teams really wasn't sure when to call us. And
since we weren't sitting there in every meeting, there was a possibility that something could come
up that might fall off the radar, that we might not get contacted for, that they may not even think
is important. So we decided that we need to give clear guidelines to the scrum masters and product
owners as to when they should include information security. That's how I created the eight simple
questions. So basically these eight questions are a guideline for that leadership team, and if they
answer yes to any of them, they need to make sure to include their information security representative
in those discussions. These are the questions as I currently have them. We're going to talk about
that in just a bit because they change over time. These are not one and done kind of things.
Am I dealing with any sensitive information? Now information that is sensitive, it's going to mean
something different for each one of our organizations. If you're working in healthcare,
if you're working with banking information, if you're working with retail with PCI information,
you will be able to define what sensitive information means for your organization.
Am I sending or receiving data outside of the company? Is there a new vendor involved?
Am I going to introduce new software systems or libraries, even open source? And remember open
source because we're going to talk about that. It's a unique challenge. Do I need access to
secrets such as passwords, keys, or certificates? Will I need the ability to use single sign-on?
Am I going to leverage the cloud? Will I need updated access roles? So these eight questions
are what we are currently using as a guideline for the interaction between the Agile team
and their information security representative. As I mentioned, the exact questions are going to
differ on your organization. The questions themselves will actually change over time.
I started I think with seven questions and then we had several issues come up with access roles
and who to ask and blah blah blah. So we added a question for that too. So don't be afraid to
take a stab at it and massage it and change it. That's the thing about Agile and DevOps,
it's changing all the time and we as information security people need to be able to change our
stuff as well. So the basic idea is this. Create a list of questions, usually 10 or less,
something that's going to fit easily on a PowerPoint slide or a single piece of paper
so they can take it to meetings with them. Give that to the Agile team leadership and they have
a guideline. They now have a concrete way of knowing when they need to work with information
security. So we found another advantage to this approach as we onboarded additional information
security people to become representatives for the Agile teams. A lot of these folks were pretty new
to the organization and they didn't really know when they needed to inject themselves into their
teams. They didn't know what they needed to listen for in the backlog refinement sessions.
So we used these 8 simple questions on our own people. It gave them guidelines and hints of what
to listen for when they're working with their Agile teams. So this kind of became a win-win
on both sides of the coin. So I think this was a great tool that we've been able to use.
Another tool that I honestly have to thank Jim Manico for. Jim, if you're out there,
thank you for this. This was a recommendation he made to me after one of our training sessions.
And that is the Security Champions Program. A Security Champions Program is a team of
application developers. We get one from each Agile team. And these developers have expressed
an interest in information security, or secure coding, or just security in general.
This is a voluntary program. The members of the team volunteer to come to the Security
Champions Program. And we typically meet quarterly. I provide them lunch when we're in the office,
thanks 2020. I look forward to getting back to the office and having lunch with these guys and
gals. They're a great team. So the team has this following charter. We assist the Agile teams in
terms of application security. We serve as a communication conduit between the Agile teams
and information security. And then we meet quarterly to share experiences, we ask questions,
we share knowledge. It has had some incredible benefits. First of all, security has a
relationship with at least one person on every Agile team. I'm not saying we have them all.
What I am saying is, sometimes we hear about things long before it percolates through the
management chain. I might get a phone call about, hey, we're thinking about using this open source
library. Is that cool? We take a quick look at it. We approve it. By the time the formal approver
gets into my hands, we've already done it and they're coding. So it gives us a great communication
conduit. And that helps me make things faster. I may have not mentioned this before. Hopefully,
I have. Relationship is key. You're going to hear about relationship throughout this
entire discussion. We build relationships with those Agile teams. It starts with having an
assigned rep. It moves into a different relationship level when we have a security
security champions team. Secure coding training takes it to yet another level. And you'll see why
in just a few minutes as we talk about the developer training. But it's about building
relationships. It's about having fun while we're doing the right thing for our organizations.
So, so far, the team has done a lot of really cool things. We've reviewed vendors and products,
both on the development side and on the information security side. We've selected
our developer training programs for the last two years based on the input from the security
champions team members. We've done a bunch of POCs together. We've cross consulted. Now,
this was kind of an unexpected benefit is where if I had a team that was very well versed in one
technology and they mentioned that in a security champions lunch, another team might reach out
and say, hey, we're about to do that. Can we sit down and talk? So, it actually starts to
cross pollinate some security knowledge and some technical knowledge between different application
teams. And then finally, it served as a candidate group for application security openings within the
InfoSec team. The software development life cycle. So, as we started to roll out security, a lot of
people have asked us, how does this fit into our SDLC? So, I ended up creating a plan that maps
information security phases to the software development life cycle as it currently exists.
It is not a perfect fit, but I think it is a good representation of how information security can
bring value to each step of the SDLC. So, this is what I like to call my dark side of the moon
diagram. Those of you old enough to remember, probably recognize the album cover. So, if you
take a look on the right, we have the software development life cycle from concept into design,
up into coding testing and QA, QA and deployment, and then post-deployment at the very top.
Now, if you look on the left side, you'll see that it matches up pretty closely with security
knowledge, security design, secure coding techniques, security testing, and then security
response. The each individual layer of the pyramid, I think, are deliverables that security
can bring that help take that concept on the left and apply it to the SDLC phase on the right. And
we're going to walk through those real quick. Training is the base of the pyramid. I think
training is foundational to everything above it. If you don't have a solid foundation of
training for your application developers, everything else is at risk. So, this particular
approach, we're going to talk about in depth, again, because it's my favorite part of what I do.
It includes things like lunch and learns, outside speakers, conferences like you're attending right
now, video training, and things like that. The next level up is threat modeling. Threat modeling,
the application team, and InfoSec, we sit together, we work together to examine the application
design, the system flows, whatever they can bring to us that they have available in that design
phase. We look, we try to work together and identify potential vulnerable vectors and try
to figure out a way that we can mitigate them. Now, threat modeling, there are books upon books
written about threat modeling. And there are people in my organization on my InfoSec team
that are much better at it than I am. And they can actually create incredible documents and spend
hours and days and days and days doing this. I think you need to start out small. I think it's
about baby steps at first. Start with a discussion. This can be a one or two hour discussion where you
have those deliverables from the design phase, and you whiteboard them. And you come out of there
with to-dos, or questions, or things to research. It doesn't have to be this big, heavyweight,
overbearing approach to threat modeling at first. It needs to be a conversation. That gets back to
relationships. You need to have those relationships in place so you can have those conversations.
The third level up is static code analysis. We're going to talk about this
in much more detail later. But most tools are designed to identify potential vulnerabilities
within the code. Now, static code analysis actually looks at the code and looks for
patterns within the code without running it. Now, we're going to see a little bit later
why that's important. I think I went the wrong way. There we go. Dynamic code analysis is the
next layer up. And that is very similar to the previous layer, which was static code analysis.
But instead of just looking at the code, the code is actually executed and tested for
vulnerabilities. So, like I said, these two pieces are very, very similar. One is looking
at patterns with the code. The other is actually exercising the code and looking for issues.
The next is pin testing. Pin testing can be performed by your internal team if you have
the skills and resources, or you can use an external firm. This is usually a greater
scope than dynamic testing. It will actually look at things not just in your application,
but system-wide for vulnerabilities and attack vectors. And usually there's a person involved.
It's not just an automated exercise. We're going to see a little bit more when we talk about
dynamic testing, the difference between that and pin testing. And then finally,
when the code is in deployment, you need to have a vulnerability management process
that you can classify, assign, and track identified issues. Right now, InfoSec meets
regularly with a representative from each application team to review these and to make
sure that they're working and they're on track. Again, those relationships are absolutely key
to making this work. So now we are getting to what I consider my absolute favorite part of my job,
and that is developer training. If you remember back to that pyramid we just looked at,
this is the foundation for everything that InfoSec does with AppDev. When I came in
to Information Security about seven years ago, I started a developer training program because I
realized that the developers really are your last best hope for defense when it comes to
application security. If your application is written correctly, then everything you throw
in front of it, application firewalls, everything like that, that's gravy. That's nice, but you've
got a really tight application, and that's where real security takes place. The problem is,
I was a developer for roughly 25 years before coming to Information Security,
and nobody taught this stuff. This is stuff you learned on your own if you had an interest in it,
and it wasn't universally available to all developers. So I started small. I started with
Lunch and Learns. It was really something that, hey, I think this is important, let's try it,
let's see if it works. And it really did. We would end up having Lunch and Learns about every other
month. I would create a class on a specific topic. My first one was cross-site scripting,
a lot of fun. It's neat when you can show people cross-site scripting examples, even if you just
run WebGoat or something like that. They get excited. They kind of dig it, and they start
trying it on their own stuff. The content usually lasts about 40 or 45 minutes, a lot like this
session, and then we leave 10 to 15 minutes for open discussion and questions. Now the key here
is, make it fun. It has to be fun. I do things like, I'll go out and I'll buy cookies or some
kind of treat, so we can all sit around, eat something together. There's usually some kind
of door prize or door prizes, and they don't have to be expensive. You can just go down to
five below and get a couple of wireless Bluetooth speakers. And people dig it. It's a lot of fun.
Again, making it fun is absolutely key to a successful Lunch and Learn program.
So once we got that up and running, I decided we needed to take this up a notch.
There is a level of training that I couldn't provide that I think our developers really,
and in order to do that, I knew that I had to reach out to industry experts.
And what we ended up doing is, we were able to get commitment from management, both on
the IT development side and on InfoSec side, that we could do one or two days where most,
if not all, of the developers will attend a training event. Now here's a hint. If you are
going to try this and you're doing Agile, put it in your Innovation Sprint. If you put it in the
Innovation Sprint, which is usually the last sprint of the quarter, and the developers have
a little more leeway of what they work on during the Innovation Sprint, you'll very likely get
more participation than if you step on a middle of a sprint where there's actual deliverables due.
So keep that in mind. I was really, really lucky. My very first interaction with this program,
I was able to secure Jim Manico, and he came in and made an incredible impression on our
development teams. Jim's been back, I think, three times now. Always a pleasure.
Rich Mogul's been by twice. We've had a recent one with Ron Paris. Dr. Philippe Des Riques,
phenomenal instructor. I just took his OAuth class, so if you're looking for OAuth,
looking forward to that. The reason I actually want to call these gentlemen out by name is I
think it's important to be grateful. I am very grateful for everything they've done for me
and my development teams. And it's important to be grateful to those who help you along the way
on this journey, because none of us make it alone. So gentlemen, thank you very much. If you
see this or hear it later, your training, your mentorship has meant a lot to me and my
development teams. So one cool aside is, I mentioned this earlier, the Security Champions
team has been involved in the selection of the instructor for the last two years.
This gives your application development teams a voice in the training. And we've seen that
participation has actually increased. We had great participation before then, but when the developers
actually had a voice in picking who's coming in and what they're going to talk about,
we're finding that it's much more applicable to what's on their roadmap. It's much more
applicable to the skills they're hoping to get. And participation goes beyond what I could have
ever hoped for. So again, we're starting to see some of these initiatives are starting to
weave together and starting to build kind of a program. It's kind of cool.
So this next one, I really wish I was doing this right now. Take a developer to DEF CON.
We had management agree to provide scholarships to developers that were selected by the information
security team. So we would give them the money to buy their badges so they wouldn't have to go and
get money out of their own bank account. So each year I've been able to bring up to three developers
to DEF CON. That's been awesome. It has been just a phenomenal experience. They've gained a ton of
knowledge and a ton of awareness by DEF CON. One of the things that I found really interesting was
I really advised them not to restrict their sessions to things that they think will apply
to their job. I did that my very first DEF CON and I wasn't really sure I was going to come back,
to be honest with you. The next year, one of my co-workers strongly encouraged me to go. Thank you,
Barry. And encouraged me to go and find a few fun things that just had nothing to do with my
immediate job. The interesting thing is they ended up having more to do with the things that were
coming down the road in my job than I ever could have imagined. So I fully encourage each one of
these developers that I bring to DEF CON to experience the full DEF CON experience. Go to
all the sessions. Try some of the capture the flags. Try whatever you want. Have a good time
because it's really important for them to be immersed in the DEF CON experience.
So like I said before, they come back with a changed perspective. It's a lot like that scene
in The Matrix where that thing crawls out of Neo and he goes, my god, that thing is real?
They understand now that the security is not just something that we're teaching and preaching about
within InfoSec, that this stuff has real world application. And it's that realization that makes
them a better developer, a more secure developer going forward. We've seen a trend where a lot of
the folks that we've taken, actually, I think all but one of the developers that we've taken
to DEF CON have eventually become security champions members. And then more recently,
in the last couple of months, one of those developers, I think he was the first one I
took to DEF CON with us, ended up joining the Information Security team on a full-time manner.
And Shane, if you're listening, welcome. I'm really glad that you're part of the team.
So there's a couple of other fun activities that we do in developer training. Again,
got to keep it fun. Everything has to be fun. You learn better when you're having a good time. I
look at DEF CON. We're all having a great time and we're learning stuff. So a couple of DEF CONs
ago, for reasons of work, I couldn't make it. It was the only DEF CON since coming over to
Information Security that I was not able to go to. And I must admit I was bummed. So what I decided
to do was I was going to have my own little DEF CON. And we decided that, okay, let's just take
people along for the ride. So we identified five videos from previous DEF CONs and we had a lunch
and learn for five days that week where we watched it. We talked about it. We had drawings for DEF
CON related swag or items or giveaways. We just kind of had a good time. It was kind of built up
as, you know, hey, take a break. Come bring your lunch. Let's hang out. Watch something from DEF
CON. Talk about it. You might win a t-shirt. You might win some stickers, whatever. And people
really enjoyed it. It was a lot of fun. As a matter of fact, I'm probably going to do one this fall
for the folks that didn't get to enjoy what you're doing right now, what you're enjoying right now.
And then finally, I have an annual thank you lunch and learn. This was a tradition I started
the first year we did lunch and learns. It's always the last lunch and learn of the year.
And I typically will go out and I'll purchase quite a few giveaways. They don't have to be
expensive. We do things for, you know, like say the Bluetooth speakers, boxes of candy. We've done
some Christmas mug sets, you know, just weird things, but things that are kind of fun.
And we do a shorter lunch and learn followed by a session where we just eat some good food
and we have some drawings. And I have an opportunity to thank my development community
for the support of the lunch and learn program and for the support they've given to information
security. I think it's important to say thank you. And this has been just an incredibly popular
lunch and learn every year. It's a lot of fun and I highly recommend it. So here are the keys to a
successful training program. Make it applicable and give the developers a voice in the training.
And usually if you give them a voice in the training, it will be applicable just because
they're going to tell you what they need. I think I've mentioned this 1427 times, but
make it fun. It's got to be fun. Share how much you care about this stuff. If you're engaged,
you're passionate about it, people are going to feel that and they're going to gravitate to you
and gravitate to that topic. And then finally build relationships. Listen, reward, and always
express gratitude along the way in this journey. So we're getting into the next phase of the
toolkit and this is what I like to call the analysis hat trick. Three types of analysis
that if we implement all three, we're going to have a much more secure code base in production.
And the first is static code analysis. You might remember that from our SDLC pyramid.
So static code analysis scans the code without running it. It looks for known patterns of
vulnerabilities. You can usually select what you scan for. Some will allow you to scan for PCI,
I almost always scan for OWASP top 10. If you're not familiar with OWASP or OWASP top 10, I
strongly recommend that you check it out. Some of the deliverables from that organization are
absolutely stellar. I've been a member for I think seven years now, just after joining the InfoSec
team. And I've leveraged a ton of their stuff. So great organization, hats off to you if you're out
there. In an ideal world, the static code analysis tools would only report the confirmed
vulnerabilities. But even if you tune the daylights out of this thing, you're still going to get false
positives. And that becomes a problem. Because originally, when we first set up this process,
I was the only one that could review the false positives and whitelist them or waive them,
as it were. That slowed things down immensely, because that was not my full-time job. That was
something else on top of my full-time job, and possibly another full-time job some days,
it felt like. So that being said, I decided to leverage the security champions. These were folks
that had some basic knowledge of security. They knew their applications better than I ever would.
And they had a passion for making secure code. Seemed like a perfect fit. So what I ended up
giving several members of the security champions team the ability to review and waive false
positives. And that has worked out great. The end result was we've had a lot more
throughput. We've been able to address false positives much more quickly. And the adoption
of the static code analysis tools have actually increased as a result of not having that delay.
So I also see static code analysis as an extension of developer training. A lot of these
tools have an IDE component. So for example, I use the SpringToolSweep Eclipse. And you can
use the plugin and execute your scan from within your developer tool. And it will take you straight
to the line that's got the problem. And in some of these tools, you can even have recommendations
pop up in the bottom of one of the windows that says, hey, you might try blank, blank, and blank
to address this vulnerability. If developers are doing this while they're developing,
not waiting for the build process, it becomes a teaching tool. Because as a developer uses this,
he or she is going to identify and learn that, oh, I need to check my input fields to make sure
they're the right type and right size. Okay. You do that enough times, it becomes second nature.
So this becomes not only a way to secure your code, but a teaching opportunity as well.
So studies have shown that static code analysis can identify up to 80% of vulnerabilities.
That's pretty awesome. If I could tell you that you can do one thing and remove 80% of
vulnerabilities, I think that'd be worth doing. So this might be something you might want to
take a look at. Here's some example products. There's three products along the top that are
commercial. OWASP has an entire list of code analysis tools at that link right there. Again,
not meant as recommendations, but these were tools that I've come across in my journeys,
and they might be worth at least acting as a starting point for your own investigation.
So the second goal in our hat trick, the second piece to the tripod, is dynamic code analysis.
This was also in the SDLC pyramid. So unlike static code analysis, dynamic code analysis
actually executes the code. Now that has a prerequisite that the code has to be compiled
and free of errors and can run before you can actually perform the dynamic code analysis.
Dynamic code analysis, I mentioned this before, is different than pen testing,
whereas dynamic code analysis looks for the flaws in the code that can be exploited.
Pen testing looks for ways to exploit the system in general, often at a greater detail,
and almost always involving a human attacker. So again, two different things. They're kind of
similar, but I really think that dynamic code analysis is really, at best, a subset of a good
penetration test. So dynamic code analysis tools can identify up to 85%, but if you combine them
with the static code analysis tools on the market, up to 95% of your application's vulnerabilities
can be identified early on. 95%, that's pretty darn impressive. So again, I'm going to throw a
few products up here for you to take a look at. I've used several of these. Again, not
recommendations, but a place for you to start. Give you a word of warning. I think I told you
earlier on, I would tell you some things that work and some things that just didn't seem to
work quite so well. Some of these tools can be weapons. I'm looking at you, Burp Suite.
You need to do things like make sure you set the scope of your testing tool so it doesn't
see that, oh, I've got a link to Facebook here. I will jump and follow that link in my crawl and I
will start attacking Facebook. Set your scope absolutely to your application only. Run this
stuff in test. Don't run it in production. If you find a SQL injection exploit and you start
to execute it, it's very, very likely that you're going to corrupt or lose data on the database on
the backside of that injection vulnerability. Also, watch out for forms. So let's say you're
doing some dynamic testing and it's automated and you're getting ready to essentially attack a web
form. Make sure you find out where that web form is going before you start your test. You can easily
overrun a service desk by submitting thousands upon thousands of emails from a web mail form.
Not saying that I've ever done that, but I know that it has happened. So the third goal in our
hat trick is open source analysis. Open source is awesome. We all use open source. Some estimates
put the percentage of open source in internally developed applications at 70%. That's 70%.
Now the problem with that is, as a result, about 50% of our applications contain one or more open
source pieces that has a high security vulnerability. And keeping that out of production is a huge,
There are two different approaches I've taken when I've tried to address this problem. The first was
a firewall slash proxy solution. There are products out there that you can actually set up that will
proxy the main repos out in the wild. And they'll check as you pull down code libraries, they'll
check them for license compliance. So you can say, hey, I'm okay with this license, this license,
but I don't want that license. So you can set that up ahead of time and it'll check each time
a module is pulled down that it is in compliance with your licensing. But more importantly,
you can also set a vulnerability threshold. So you could, for example, say, hey, I don't want
any critical or high translation. I don't want any CVE 7 or higher coming into my organization's
local repository. So the problem with that is a lot of the tools that developers use,
particularly in DevOps for automating building of systems, automating testing,
have vulnerabilities because they really weren't meant to go live anywhere. They were basically
developer aids, developer tools, and your policy is going to quarantine or prevent them from coming
into your local repository. The result is you get a lot of phone calls. Hey, this module, I need it,
I need it now, and it's quarantined. Well, that requires getting out on the firewall or proxy
and looking at what's quarantined, determining how it's going to be used, where it's going to be used,
and does that CVE apply in this use case, and then deciding whether you can waive it or not.
It is a labor-intensive process. I didn't really like doing it this way at first,
so I ended up moving to a build-based solution. Build-based solutions are very similar to what
we're going to see in just a few minutes, where you basically add an open source review step to
your automated build process. If you find that same vulnerability as it's going to production,
let's say it's a CVE of 8, and it's going into production, you fail the build. At that point,
the developers have the opportunity to go, okay, I need to get a security waiver, or I need to find
a more recent or more updated library that doesn't have that CVE as an outstanding issue.
So this ended up reducing the manual intervention. The only problem is you could end up having
something on your internal network with a CVE of 7 or higher that's a developer tool. You've got
to be very careful when you're whitelisting or removing items from quarantine to understand how
it's going to be used, and if the CVE itself can be exploited in that manner. So here's some
products once again. OWASP, I've mentioned them several times. Great organization. Again,
these are not recommendations. These are simply places for you to start if you haven't
looked at these already. And finally, let's talk about automation. So those last three steps,
they seem like a lot. And originally we were doing them manually, and it became unwieldy
very quickly. It just took a lot of time. It took a lot of effort. So what we were able to
do was automate most of this into the build pipeline. So if you have a continuous integration
tool, something like Jenkins, you can add steps that execute static code analysis, dynamic code
analysis, and open source analysis with every build. So this gives you all of that protection
over 95% from the studies with every single build that you do. Now, there's some things that are
just too stinking big. So if you're still dealing with legacy applications, big old monolithic jars
and ears, if you've got something that's somewhere in the hundreds of thousands of lines of code,
sometimes close to a million lines of code, this is not going to go well for you because
the dynamic scan, the static scan, those pieces are going to take a long, long time.
I tested one application for a friend of mine just to try it. It was a legacy application,
and the scan for just the static code analysis piece took over 12 hours. You can't stop a build
for 12 hours and expect the developers not to be concerned. So I don't recommend this
for large applications. If you do, you might do it as an asynchronous build step. So basically,
it says, hey, I've got to go do static code analysis. I'm going to launch it and I'm going
to continue with my build. You lose the advantage of being able to block at that point, but you
still have an automated scan going along. You'll need to have a process in place to alert the
development team as well as information security of the results of that scan, and then decide how
you want to react. But you lose the ability to block. But fortunately, the industry is really
moving towards smaller builds, APIs, microservices, things like that. And inline scans are very,
very possible. They're very fast. These tools are tuned for this kind of work,
and they run incredibly, incredibly fast. Just aside, some of the tools
are still working on their automated pipeline plugins. So if you are looking at any of the
tools, whether it's the ones I've mentioned or something else, make sure that you ask about
the maturity of their continuous integration pipeline plugins.
Otherwise, you might find yourself stuck manually doing this for a while until you can find a way
to do that. I want to end with a couple of final thoughts. So the keys to success, hopefully you
can read this, they are start small. Do one thing. For me, it was starting with a training program
that literally led to everything else that you saw in this presentation. And as you saw in the
presentation, this stuff tended to feed upon itself. The developer training led to more
developer training, which led to a security program. So start small, and it will all come
together. Get management buy-in by both Information Security and by AppSec Management.
The centerpiece right there, that has to be the key for me. It has to be fun. When this stuff is
no longer fun, I'm not going to do it anymore. We've got to keep making it fun, particularly for
our developers. Developers are doing a lot of stuff. Adding security on top of everything else
they're doing is a pain in the butt. And for the most part, it's not welcome, unless you can make
it fun. If you can make it fun, it's going to happen. So just like we just mentioned, automate
everything you possibly can. And remember, it's all about those relationships. The relationships
with your scrum teams. It's the relationships with your developers. It's the relationships
with your vendors and the trainers that you can get to come in. So building a good relationship
is key. Don't try to reinvent the wheel. The best innovation is usually stolen from somewhere,
or given from somewhere. So if you can take one piece from this discussion and implement it in
your organization, I will feel blessed, and I'd love to hear about it. Take it slow. This stuff
is new for all of us, and it's always changing. And finally, learn what works best in your
environment. So I want to wrap up with thank yous. I want to thank you for spending the last 40,
45 minutes, whatever it ended up being, with me and with AppSec Village. I want to thank AppSec
Village. I got to volunteer there last year. I ended up spending the entire DEF CON there.
After, I think that was my sixth year in information security, I finally found my home,
and that is the AppSec Village. So thank you to the leadership of AppSec Village for doing this,
and for giving me this opportunity. I want to reiterate my thanks to Jim Manico, and he's kind
of been my spirit leader on this journey, and I appreciate all you've done. My colleagues Barry
and Shane, my partners in crime, you guys are awesome. And finally, my best friend, my wife,
who's put up with me doing this and several other things in information security. She's absolutely
the best, and I can't thank her enough. So thank you guys very much. I appreciate you spending this
time with me, and I hope you have a really, really excellent rest of your DEF CON. Thanks.