Here you will find frequently asked questions about BSIMM. For the full and unexpurgated model, download the BSIMM report here.
BSIMM (pronounced “bee simm”) is short for Building Security In Maturity Model. BSIMM is a study of real-world software security initiatives organized so that you can determine where you stand with your software security initiative and how to evolve your efforts over time.
Software security is about building software to be secure even when it’s under attack. As we’ve learned from years of reviewing network security breaches, protecting software is much easier if the software is built with security in mind. Furthermore, security is a property and not a thing, so software security—being resistant to attack—involves much more than simply adding security features like encryption or passwords to software.
Organizations that develop and depend on software to do business (and everybody does these days) need software that won’t leak millions of identity records, call election results into question, incur huge legal liabilities, or allow secrets to fall into the wrong hands. The only way to make software trustworthy is to build security in. In short, everyone who relies on software needs BSIMM.
We built BSIMM entirely from observations we made by studying real software security initiatives. BSIMM does not tell you what you should do; instead, it tells you what everyone else is actually doing. This “observe and report” approach to software security science stands in sharp contrast to prescriptive approaches based on personal experience.
There are 128 firms included in BSIMM12. On average, they had practiced software security for 4.4 years at the time of their current assessment (with values ranging from less than a year to 16 years as of September 2021). All 128 firms agree that the success of their initiatives hinges on having an internal group devoted to software security—a software security group (SSG). The average size of an SSG is 22.2 people (the smallest is 1, the largest is 892, and the median is 7.0). Often there is a satellite group of others (developers, architects, and people in the organization directly engaged in and promoting software security), and that group on average consists of 50.4 people (the smallest is 0, the largest is 1,500, and the median is 1). The average number of developers in participating organizations is 3,113.6 (the smallest is 5, the largest is 100,000, and the median is 850), yielding an average ratio of SSG to development of 2.59% (the median is 0.74%).
All told, BSIMM describes the work of 9,285 SSG members and satellite staff working together to secure software that powers—nearly 153,519 applications—and is built by 398,544 developers.
The executives in charge of the software security initiatives we studied have a variety of titles. Examples include:
We observe a fairly wide variety in terms of where the SSG is situated in an organization. In particular, 67 of the 128 participating firms have SSGs that are run by a CISO or report to a CISO as their nearest senior executive. And 18 of the firms report to a CTO as their closest senior executive; 4 report to a CIO, 11 to a CSO, 3 to a COO, 2 to a CRO, and 1 to a CAO. There are 19 SSGs that report through some type of technology or product organization.
For each activity, we give a description and one or more real examples to illustrate how organizations make it happen. The examples are never the only way to conduct a given activity, but we think they’re helpful for understanding software security reality.
Don’t worry—BSIMM is an observational model, which means that when we see an activity being conducted in multiple participant organizations, we add it to the model. The model is cumulative, and no organization carries out all activities. Over the years, we’ve found a surprising amount of common ground between the financial services organizations, independent software vendors (ISVs), and IoT companies we studied, but their initiatives are by no means identical, and every organization is at least a little bit different. You wouldn’t implement a direct copy of a friend’s financial plan, so you shouldn’t expect to lift someone else’s software security initiative either. Use BSIMM as a source of ideas and general guidance—as a trail guide rather than as a cookbook.
The 12 highlighted activities are those we observed most often in each practice.
|[SM1.4]||Implement lifecycle instrumentation and use to define governance.|
|[CP1.2]||Identify PII obligations.|
|[T1.1]||Conduct software security awareness training.|
|[AM1.2]||Create a data classification scheme and inventory.|
|[SFD1.1]||Integrate and deliver security features.|
|[SR1.3]||Translate compliance constraints to requirements.|
|[AA1.1]||Perform security feature review.|
|[CR1.4]||Use automated tools.|
|[ST1.1]||Ensure QA performs edge/boundary value condition testing.|
|[PT1.1]||Use external penetration testers to find problems.|
|[SE1.2]||Ensure host and network security basics are in place.|
|[CMVM1.1]||Create or interface with incident response.|
Three verticals in BSIMM operate in highly regulated industries: insurance, healthcare, and financial services. In our experience, large financial services firms reacted first to the regulatory changes of the 1990s and early 2000s and started their SSIs much earlier than insurance and health care firms. Even as the number of financial services firms doubled over the past five years (adding a large influx into the BSIMM data pool of newly started initiatives), the average age of financial services SSG at assessment time remains 5.4 years, versus 4.4 years for insurance and 4.2 years for healthcare.
Time spent maturing their collective SSIs shows up clearly in the side-by-side comparison. Although the insurance vertical includes some mature outliers, the data for these three regulated verticals show insurance generally lags behind in software security. We see a starker contrast in health care, with virtually no outliers.
BSIMM is not a standard like ISO 27001 or the official rules of table tennis. Instead, BSIMM describes the set of activities practiced by the most successful software security initiatives in the world. In that sense, it is a de facto standard because it’s what organizations actually do. You could say we discovered it rather than invented it.
If you don’t have a software security initiative, you need one. And you can use BSIMM to get started: It can help you figure out how many people you’ll need in your software security group, what those people should do first, and what kinds of things they’ll probably be thinking about in a few years. If you already have a software security initiative, you can use BSIMM to learn where you stand and make plans for the future.
BSIMM is free: We released it under a Creative Commons license. This means it’s as “open” as any other model, and you can take it and use it as inspiration for your own internal documents, or use our published data to make a model of your own. If you do those things, you’re required to tell people where the material came from. In other words, point back to BSIMM. If you need a little help, contact us.
All BSIMM participants have an internal group devoted to software security—the software security group (SSG). We’ve never observed an organization carrying out the activities in BSIMM successfully without an SSG. We noted an average of 2.59%, across the 128 organizations we studied. That means one SSG member for every 39 developers when we average the ratios for each participating firm. For organizations with 500 developers or fewer, the largest ratio observed was 51.4% and the smallest was 0.33%. For organizations with more than 500 developers, the largest ratio observed was 14.9% and the smallest was 0.08%. To remind you of the particulars in terms of actual bodies, SSG size on average among the 128 firms is 22.2 people (the smallest is 1, the largest is 892, and the median is 7.0).
For many software makers (including ISVs, banks, health care firms, governments, and others), software security has been a twenty-first-century concern at best, and an executive-level concern for 10 years at most. The collective “we” are just now reaching the point where we’ve accumulated enough experience to compare notes and talk about what works at a macro level. Secure programming, penetration testing, and the like have been topics for a while now, but the best methods for organizing software security initiatives have taken longer to emerge. BSIMM captures those activities into an observational model that is free and open for everyone to use.
High-profile ransomware and software supply chain disruptions drive increased attention on software security
Over the past two years, BSIMM data shows a 61% increase in the “identify open source” activity and a 57% increase in the “create SLA boilerplate” activity among participant organizations.
Businesses are learning how to translate risk into numbers
Organizations are exerting more effort to collect and publish their software security initiative data, demonstrated by a 30% increase of the “publish data about software security internally” activity over the past 24 months.
Increased capabilities for cloud security
Increased executive attention, likely combined with engineering-driven efforts, has also resulted in organizations developing their own capabilities for managing cloud security and evaluating their shared responsibility models. There was an average of 36 new observations over the past two years across activities typically related to cloud security.
Security teams are lending resources, staff, and knowledge to DevOps practices
BSIMM data shows a shift by software security groups away from mandating software security behaviors to a partnership role— providing resources, staff, and knowledge to DevOps practices with the objective to include security efforts in the critical path for software delivery.
Continuous defect discovery and continuous improvement
BSIMM12 data indicates that more firms are implementing modern defect discovery approaches and favoring continuous monitoring and reporting rather than using a point-in-time defect discovery approach.
The new cadence is driving the need to provide data to leadership to support governance decisions, as shown in the 30% growth of the BSIMM activity “publish data about software security internally” over the past two years. While governance processes remain mostly manual today, organizations are trending toward governance-as-code, observed in 15% of the firms measured for BSIMM12.
Security testing in QA automation more than doubled
The observation rate of the BSIMM activity “integrate opaque-box security tools into the QA process” increased by more than 50% over the past two years. Similarly, the observation rate of the activity “include security tests in QA automation” also more than doubled over the past two years.
Software Bill of Materials activities increased by 367%
BSIMM data shows an increase in capabilities focused on inventorying software such as creating a software Bill of Materials (BOM); understanding how the software was built, configured, and deployed; and increasing an organization’s ability to redeploy based on security telemetry.
Demonstrating that many organizations have taken to heart the need for a comprehensive up-to-date software BOM, the BSIMM activity related to those capabilities—“enhance application inventory with operations bill of materials”—increased from 3 to 14 observations over the past two years—a 367% increase.
“Shift left” progresses to “shift everywhere”
“Shift left” focuses on moving security testing earlier in the development process. “Shift everywhere” extends the idea to making security testing continuous throughout the software life cycle, including smaller, faster, pipeline-driven security tests conducted at the earliest opportunity, whether in design or production.
The move away from maintaining traditional operational inventories and toward automated asset discovery and creating Bills of Material includes adding “shift everywhere” activities such as using containers to enforce security controls, orchestration, and scanning infrastructure-as-code. Increased BSIMM observation rates of activities such as “enhance application inventory with operations Bill of Materials,” “use orchestration for containers and virtualized environments,” and “monitor automated asset creation” all demonstrate this trend.