- Cybersecurity Ecosystem Show
- Posts
- Patching, Threat Modeling, and M&A Due Diligence: Lessons From Security Veteran Brian Harden
Patching, Threat Modeling, and M&A Due Diligence: Lessons From Security Veteran Brian Harden
From the Cybersecurity Ecosystem Show conversation with Brian Harden, information security leader and longtime hacker

-
Brian Harden has worked in information security for 30 years, with experience at Microsoft, IBM, and DEF CON. He describes himself as a hacker, and he has held the same view of the industry's biggest gap for most of his career: organizations still do not patch reliably, and they still skip the design review that would prevent many bugs from shipping.
On this episode of the Cybersecurity Ecosystem Show, he explained why patching has become harder, what a working vulnerability management program requires, how to get other departments to cooperate, what he checks during an acquisition, and how people can get into the field today. He closed with the story of his push to get Keybase to add basic privacy controls.
Here are the main lessons from the conversation.
Why is patching still a problem in 2026?
Brian's answer to what is broken in the industry was immediate. "Why in the year of our Lord 2026 are we still screaming at people to patch?" he said. "I think I had an easier time getting people to patch in the 90s than today."
The tooling has improved, so the cause is elsewhere. His first explanation is dependencies. Software used to rely on a few libraries. Today a product pulls in third-party packages that have their own third-party packages, and by the time it ships there can be dozens of them underneath. A vulnerability is often in something the code depends on, and the attack surface is much wider as a result.
At one company, his team ran a full count of vulnerabilities in the environment and found thousands. Many were not exploitable, and some would have given an attacker nothing of value. The volume still shapes behavior. The accepted practice has become fixing criticals, trying to fix highs, and planning to reach mediums and lows later.
"We never get to the mediums and lows later," Brian said, "because later never comes."
His second explanation is that the basics do not get attention. "The basics are boring," he said. "Patching's like eating your vegetables. I don't wanna, but I know I need to."
Why do organizations avoid patching?
Brian described a cycle. Companies hold off on patching until it becomes a fire drill. The fire drill is painful, so the team is reluctant to patch again, and the next round waits for the next emergency. The avoidance produces the pain the team was trying to avoid.
Under that cycle is fear of breaking production. "The big fear with any operations person is: if I touch this, what happens if it doesn't come back?" he said. "Every business says, we move fast and we break things, but at the same time there's also this undertone of, and you better not break things."
He pointed to Equifax as the example of what delay costs: a simple fix that kept getting pushed back.
Legacy environments add a knowledge problem. After acquisitions and staff turnover, nobody may fully understand how the stack works. A 20-year-old code base may have been written by someone who left 15 years ago, and the documentation may exist only in that person's memory. "If I didn't build it, I don't want to touch it," Brian said. "That's where the core of the problem is."
How do you build a vulnerability management program?
Brian laid out four requirements.
Know what you have. "It's amazing how many companies out there literally don't know what they have," he said. Many companies are breached through a system they did not know existed. He named runZero (formerly Rumble) as an asset discovery tool he rates highly.
Track your vulnerabilities. Use software that records which vulnerabilities exist in the environment.
Rank them by risk. Time and staff are limited, so patching has to go where the exposure is highest: internet-facing systems, and systems that touch, transmit, or hold personally identifiable or financial information. "If I can only patch vulnerability A or vulnerability B, which do I patch and why?" He sees this prioritization as a strong use for AI in security.
Set a cadence. Weekly, monthly, or quarterly, the schedule matters less than having one and planning for it. That includes deciding what will be patched and what the rollback plan is. Scheduled patching replaces the out-of-band, panicked version that makes teams want to stop.
His summary: "It's a thing you will have to do until the heat death of the universe. It's time to come to peace with it and figure out how do I do it in the least painful way."
How do you get IT and other teams to support patching?
Brian runs every request through one question from the other team's point of view: "What do I win by playing?" If security asks IT for heavy work that pulls people off their projects and gives them nothing in return, IT will decline. The request has to be presented in terms of what the other team gains.
The second requirement is regular contact with peers before anything goes wrong. Managers at several of his employers have told him that people value the time he takes to talk with them. His reason is direct: "I need them to be successful, just like they need me to be successful."
"You don't want to build these relationships during times of crisis," he said. If the first real meeting between the head of security and the head of IT is at two in the morning because the front end went down, something was missed earlier.
Regular contact also allows scheduling to be negotiated. If security proposes the first Thursday of every month and IT already has maintenance that day, the person asking needs to be flexible. Brian has also worked incidents where IT blamed security, security blamed IT, and the cause turned out to be a deployment. If the teams had known about each other's work that night, most of the people on the call could have gone back to bed.
Why should you threat model early and often?
"The cheapest bug fix you will make," Brian said, "is reviewing things ahead of time and not creating the bug." He sees the push to release a minimum viable product as the reason companies skip that review and then go straight into response mode after launch.
He gave an example from a software company where the product was a set of smaller products working together. He asked one developer to walk through his service: where the data comes from, what the service does to it, and where it goes afterward. The developer drew it on a whiteboard and kept correcting himself as he remembered where the build had changed from the plan. At the end he asked Brian to take a picture, because it was the most accurate depiction of the service that existed.
The larger finding came from the group. "Every single person thought they knew what their related components did and how they worked, and they were wrong a hundred percent of the time," Brian said.
Threat modeling with everyone in the room corrects those assumptions, and it is the right time to set security requirements. Telling a team during design that certain fields need encryption results in a note. Telling them a week before launch that half the tables need encryption results in a missed date.
He also recommends a second threat model before go-live. "Nine times out of ten, how you ended up building it was not how you originally intended to build it."
What does "hacker" mean?
Brian's definition has stayed the same throughout his career. "It's understanding how things work and getting them to work in ways that weren't intended," he said. That includes getting past limits, whether they were placed on purpose or are just limits of the product.
He described the mindset with an example: a form asks for a full name, and the hacker wonders what happens if it receives 7,000 Santa emojis instead. "You look at limitations and boundaries as suggestions," he said. He added that any experienced hacker has at least one story about breaking something by accident.
How do you get into cybersecurity today?
Brian mentors often, and he tells people his own path is no longer available. "I started doing information security and an industry built itself around me," he said. His current advice has four parts.
Narrow your focus. He gives mentees a list of information security domains and asks for a ranked top ten. Then he has them draw a line under the third item and ignore everything below it. The field now supports specialists, so the first step is choosing what you want to do.
Match certifications to that focus. Strong penetration testing certifications exist, but they are a poor investment for someone who wants to work in policy or GRC.
Build a real network. "InfoSec has always been a who you know world," he said. In 30 years he has received two job offers from applying online without a contact. Every other role came through someone he knew. He recommends ISSA and ISACA chapters, local hacker meetups, DEF CON groups, and 2600 meetings. He also warns against showing up only to extract something. Security people assess risk for a living and notice it. New entrants also tend to group together, which does not help because they are all looking for the same opening. "Find the veterans, talk to them, make friends with them."
Get more than one mentor. A technical mentor and a business mentor are usually different people. Brian spent years as an engineer refusing to engage with company politics and kept running into walls. "You're playing the game whether you like it or not," he said. An early mentor of his put it this way: "Men of action are ultimately controlled by men of thought."
For people already working in IT, networking, or software development, his advice is to get to know the company's security team. He has seen many people move over that way, and security teams want them. A developer who moves into application security already knows how to talk to developers.
What should security check during mergers and acquisitions?
Brian has worked both sides of acquisitions, mostly as the acquirer, including Microsoft's purchase of Skype. His rule is "trust the data, don't trust the story."
Both sides have reasons to want the deal closed. The people being acquired stand to make money, and the internal sponsors may have their careers tied to it. That can lead everyone to overlook problems. So he asks for evidence:
The information security program itself
How assets are identified and managed
How vulnerabilities are managed
The code security pipeline. "Show me how bits get from a developer's keyboard to sitting on a server that a human being interacts with."
Third-party attestations, including whether a penetration test was done in the last quarter and whether a reputable firm did it
He then talks to employees. Outside the executive team, most people at the acquired company are afraid of losing their jobs. Brian's guidance is to be as transparent as possible and to be honest about what you do not know. During the Skype acquisition, employees asked him whether layoffs were coming. He told them that as far as he knew the answer was no, and that he was not in those meetings either. In his experience people prefer an honest "I don't know" to "you'll be fine."
Those employees are also the ones who know how the business runs, and they sometimes disclose problems that are not in the documents, such as a log server full of PII. Brian said he has never seen a deal canceled over a security finding, but he has seen deals delayed until a serious one was fixed.
For security leaders on the acquired side, his advice is to ask to be involved and to meet the acquirer's security team early, to learn whether they are experienced practitioners or will judge the program from a spreadsheet.
What the Keybase story teaches about listening to security feedback
Brian was an early Keybase user because it made encryption easy and let users hold their own keys. In his account, the problems began after the company aligned itself with the cryptocurrency community. Anyone could message any user or add them to a group. Users received scam messages and requests for tokens, large group chats were used to sell get-rich courses and push malicious links, and women on the platform received unsolicited explicit images. "The encryption was there, but there was no privacy," he said.
He tracked down the founders, who were difficult to reach, and asked for basic controls: an option that only contacts can message a user or add them to a group, set as the default. He said the initial response was silence and then dismissal. He and other affected users kept raising the issue publicly, a technology writer covered it, and the coverage spread. Keybase then added the controls.
His view is that the company reached the right outcome for the wrong reason and too late. Keybase was later acquired, and he believes the delay on spam and harassment is a large part of why people stopped using it. "By the time they fixed it, the damage was done."
The lesson he draws applies to any company: "When security minded people come to you with things, even if you don't take action, at least listen to them and evaluate what they're saying." A reply saying the issue was on the roadmap for the next quarter would have been enough for him.
The basics are the program
Brian's points share a theme. Most of the avoidable pain in security comes from skipping work that is well understood: patching on a schedule, reviewing designs before building, checking evidence before a deal closes, and maintaining relationships before an incident. None of it is new, and all of it is still inconsistently done.
The full conversation is worth your time. You can reach Brian on LinkedIn by searching for Brian Harden in the Nashville area.
Here are the key takeaways from this blog:
Patching is harder because of dependencies: modern products rely on dozens of third-party packages, and most organizations fix criticals and highs while mediums and lows wait indefinitely.
Avoidance creates the fire drill: teams delay patching because it hurts, and the delay guarantees the next emergency.
A vulnerability program has four parts: asset inventory, vulnerability tracking, risk-based ranking, and a fixed cadence with a rollback plan.
Cooperation depends on shared benefit: present requests in terms of what the other team gains, and build peer relationships before an incident.
Threat model at design and again before launch: teams misunderstand neighboring components, and security requirements cost far less early.
Careers are built on focus and relationships: choose three domains, match certifications to them, build real connections, and find both a technical and a business mentor.
In M&A, check the evidence: ask for the program, the pipeline, and recent third-party testing, and talk to employees below the executive level.
Listen when security people raise an issue: Keybase fixed its privacy controls only after public pressure, and by then users had left.
The Cybersecurity Ecosystem Show connects practitioners, investors, vendors, regulators, and everyone in between. New episodes drop weekly. Subscribe so you don't miss one.
Reply