
Confident Code for Confident People: How Do You Know What’s Running In Your Environment?
There’s never a bad time for a RANT roundtable, but certainly, gathering together a group of highly experienced security leaders in the days following the announcement of the PHP2Shell vulnerabilities did rather imply that this one may be tricky to keep on topic. Fortuitously, the theme selected weeks previously proved entirely apt. But an entertaining early accident – when the venue staff ushered in a latecomer who sat at one of the two empty places at the table, wondering why they couldn’t find their name on any of the place settings, only to belatedly realise they’d been brought into the wrong private dining room – inadvertently helped keep minds focused on the task in hand.
How can businesses be sure that they’re not allowing third-party software to create ways through their security that they don’t know about? What steps can the modern business take to close loopholes attackers could exploit in code bases they not only have no control over, but may not even realise are deployed in their environments, because they’re there as part of a solution supplied by a third party under a different brand name? And, of course – when something goes wrong (because it’s always “when” these days, not “if”), who’s legally and financially responsible?
“I’ve been in cybersecurity for three decades, going back to the days of anti-virus,” Matt Gyde, chief revenue officer of event hosts ReversingLabs, said in his scene-setting opening remarks. “Then we evolved into perimeter defence, firewalls, blocking; then went to defence-in-depth, and then to zero-trust, and authenticating everything. But in that whole time, no-one’s looked at software differently. With the SolarWinds incident, the compiler was compromised, so people could use the platform to compromise organisations. PHP is similar. Interesting things continue to happen. So that’s why we’re talking tonight about whether the software you buy is safe to deploy and ready to go.”
It was, of course, a massive topic, and needed some expert direction. RANT’s guest host for the evening, Adam Le Tissier – UK-based vice-president for cyber resilience and incident response at the Japanese investment bank Nomura – suggested some priorities for the ensuing hour’s conflab.
“At the code level, what regulations do suppliers have to secure their code to – and what are they doing to secure it?” he began. “Today we’re seeing a lot of threats that aren’t targeting organisations directly, and to some extent we’re all third parties. So we need to know how secure are those supply chains? And, if you bring this stuff in house, it’s about looking at your own code, and how you manage those same issues. Is this policy driven, or is this security driven? These are the questions we need to look at.”
Try Your Luck
First, some context. Why are we all in this mess to start with? And how bad is it, really?
One thing the PHP disclosure had brought home to many – and, it was noted, by no means just to cybersecurity specialists – is the nature of the modern enterprise’s software estate. A product is selected from among competitive tenders solicited following a rigorous and diligent requirement-drafting process, and is deployed in the environment as a single, standalone, coherent tool. Yet chances are that significant chunks of the code that run it are third-party off-the-shelf products folded in to perform specific functions.
That’s fine, so long as everyone knows about it – customer and vendor alike. But problems start to arise the longer-established – and therefore the more successful, and more widely used – a suppliers’ product is. And when the product is not just part of the general functionality of the buyers’ network, but has been bought to fulfil critical security tasks, the problems become more pronounced.
“If you look at long-term security vendors, they’ve had nine generations of developers come through, so we’re abstracted now from the original source code,” Gyde said. “That abstraction layer is creating even more problems with the supply chain because we don’t know what was originally there.”
Having a rigorous due-diligence process in place is essential, Le Tissier argued.
“Especially when using AI,” he added. “What are they doing with it, and what controls are in place? Also, there may well be code sitting there that developers are building on top of – and maybe there’s passwords in there, and…” He tailed off, a rueful smile hinting at the potential for exasperation that quickly attends any such attempt to find out what code – and therefore what vulnerabilities – may be hiding somewhere inside the enterprise.
Gyde noted that this was exactly the problem that the PHP disclosure had laid bare. Some organisations were using versions of the software that were 10 or 12 years old, and are no longer being supported. To make matters worse, he added, even if a diligent customer organisation can persuade a vendor to give them a bill of materials, that will only tell the customer what the supplier knows is inside its code base. Little wonder that security leaders have limited – if any – confidence in the information that is supplied.
“You ask them about what’s in the code base, and what they do down in their supply chain,” one CISO sighed. “But deep down you know there’s a lot of…” They paused, thinking carefully about their next words. “‘Lies’ is too strong a word,” they eventually decided. “But let’s call it what it is: a strategy of hope.”
Control
ReversingLabs’ approach is to scan binaries, ensuring that organisations can examine the code they rely on at its component level, irrespective of how it’s badged or identified. This gets around the problem of a supplier failing to disclose (or perhaps not even realising) what earlier software their product incorporates. The firm’s products can scan PHP, among innumerable software products and formats. This will help surface any areas of concern – but it is not a solution in and of itself.
Take, for example, the situation laid out by one security leader, whose enterprise relies on Microsoft products. Their concern wasn’t so much with finding new vulnerabilities, but over how they, as an entity, can demonstrate – internally, to the board; externally, to regulators and major customers – that their systems can be relied upon.
“It’s so hard at the moment,” they said. “Microsoft say they can do this, but we get sent loads of patches every month.”
“The way I look at it is to treat all software as untrusted,” another CISO said. “So to me, it all comes down to zero trust, particularly in the post-Mythos era. Everything has vulnerabilities, so I’m going to limit access to it all. How,” they then asked, politedly but pointedly, “can ReversingLabs give me so much confidence that I no longer need to do that?”
“We can discover things based on the code the machine is reading,” Gyde said. “When you talk about agentic software, we’re actually selling more to machines today than to humans. We have about 422 billion items of goodware and badware [that the system recognises].”
“So, you can show that everything across my environment is riddled with vulnerabilities…” suggested another security leader, with a somewhat mischievous look on their face.
“We’ll say you’ve got problems, for sure,” Gyde said, acknowledging the friendly jab. “We’re not saying zero trust isn’t a model you should follow. But we can give you visibility to enable you to make controls for that software. ‘You’re running version 2.0.1, but if you swap it for 2.0.3, that’ll solve your problem’ – we can give you that kind of visibility. We can give you the detail: it’s up to you how you action it.”
Break It Bought It
Another line of questioning arose at this point. How, if at all, has “the marketing campaign called Mythos” affected potential customers’ views of ReversingLabs, one CISO asked. Their phrasing – greeted with wry chuckles around the table – reflected the widespread view that much of the hair-on-fire hysteria surrounding generative so-called AI systems’ supposed ability to find flaws in deployed software is less a realistic response to game-changing new defensive capabilities, and more a return to the bad old days of FUD (fear, uncertainty, doubt) which was the cornerstone of the PR pitches from a raft of companies – many no longer in business – who would happily try to flog their spiffily-presented but only partially effective security tools by trying to generate panic among their customer base. Are companies – this CISO wanted to know – saying, “I don’t need ReversingLabs, because Mythos will find all the vulnerabilities?”
“We’ve got an organisation we work with – they’re more into tick-box compliance – that are saying, ‘Mythos can solve all our problems’. Those are hard discussions to have,” Gyde admitted. “I don’t see Mythos as a security product,” he continued. “We add a tremendous amount of context, and we can provide the verdict to the AI. Personally, I think it’s a boon for us. We don’t really care about what [security or discovery tools] you’re putting in – we’re more concerned about the context.”
Have any customers – particularly end-user entities as opposed to code suppliers – mandated their software vendors to use ReversingLabs, another CISO asked.
“One big financial organisation based in the U.S. is doing exactly that,” Gyde said. “They say to the vendor, ‘Either I’m going to scan it, or you can do it before’.”
A couple of leaders around the table noted that you’d have to be a pretty big customer to have sufficient leverage to force any supplier into this – while accepting that a significantly sized bank would probably fit that bill. So it won’t help everyone, though might be a useful tactic for some. Still, another CISO noted, insisting a supplier scanned the code before shipping it would only help if both customer and vendor were in perfect alignment over what needed to be checked for, and what constituted an acceptable result. Again, the concept seems sound, but detailed customer context is all-important. Or, as one leader put it, if you’re trusting a third party to give you assurance, “how do you have confidence in your confidence?”
Who Knows What You’ll Find?
The final segment of the discussion coupled key themes explored earlier with questions around accountability, responsibility and regulatory approval. While there may only be some customer organisations who are covered by statutory regulation, all organisations of any size whatsoever have to reckon with broader regulatory responsibilities. While a bank or a power supplier will be accountable to regulators with significant audit powers, even a one-person business selling stuff through a website and working from home will have to comply with the General Data Protection Regulation. If the PHP furore has done anything here, it’s perhaps helped to get the message across that cybersecurity isn’t just something that big firms with temptingly large financial resources and customer databases need to worry about.
“My wife rang me in a panic over PHP, saying ‘I need to do something about my website’,” one security leader said. “The response to this from the non-cybersecurity world has been orders of magnitude greater [than to previous high-profile cybersecurity stories]. We have to take that as a win: the world has moved on from, ‘I’m too small, who’s going to hack me?'”
“Most of our exposure to it was in other software,” another leader acknowledged.
“It’s similar to SSL libraries,” said another. “It’s critical software, but the vendor has no interest in updating it…”
“…or the vendor no longer exists,” a fourth added.
“When the auditor shows up, they’re always asking, ‘Have you gone into the libraries, and further down?'” another CISO noted.
“I completely agree,” said another. “People either hide it, or say it doesn’t exist until it becomes an issue.”
Regulators and auditors, of course, are unlikely to look particularly kindly on situations such as this. Business leaders may have legitimate reasons not to go poking around at the binary level and finding vulnerabilities that they didn’t otherwise know that they had if they’re about to head into a series of conversations with regulators. But dealing with them need not be an entirely adversarial process, one CISO suggested. Also, they argued, the kinds of interactions and conversations organisations have with regulators will be different in nuance and scope to those they have with auditors.
“Regulators need assurance that you understand your risks and are managing them properly, and to know what you’re deploying in your environment; auditors check that you’re doing it all – ‘You have this policy, are you following it?'” they said. “That means conversations with regulators and auditors are different. But those conversations are not changing.”
“It’s also about making sure people understand it’s just one part of the security posture,” another CISO added. “It’s too easy to get diverted by the elephant in the room and miss the little mouse that goes and chews through the cables. The problem is that you’ll find something if you go looking for it.”