This is why high-assurance security products were/are required to have:
1. Clear description of every feature and requirement in system.
2. Mathematical spec of those where English ambiguity could effect results.
3. High level design with components that map 1-to-1 to those.
4. Low-level, simple, modular code mapping to that.
5. Source-to-object code verification or ability to generate from source on-site.
What people in faux security mocked as mere "paperwork" or "red tape" were actually pre-requisites for defeating subversion my mentally understanding a system from requirements all the way to code. A problem like this would've been impossible in such a system because it would be beyond obvious and probably unjustifiable with requirements tracing.
Every story like this further validates the methods that consistently produced systems without all the security problems plaguing modern security products. Situation isnt inevitable or even necessary: merely an inversion of scientific method where security companies and professionals consistently refuse to use what's proven to help and reuse tactics proven to fail. It's gotta stop.
That it wont is why I favor liability legislation tied to a reasonable baseline of practices. We can use an inexpensive subset of what worked in highly assured systems. 80/20 rule. Baseline would look more like Secure64 or HYDRA firewall than shit like Fortinet and Juniper. Hackers would work for exploits. I know Im dreaming, though, as DOD and NSA just dropped mandate to EAL1 w/ 90 day review for some stuff. (Rolls eyes).
Your assurances are good but the software simply _must_ be open source with a reproducible build.
We're kidding ourselves to put our faith into these closed-source products and that's only just now becoming clear. Open source. It's the only thing that will work for us long term.
Did you read my comment at all before replying? One step is the source, one step implies reproducible build, and alternately specifies a stronger requirememt (source-to-object verification).
And no, OSS with reproducible builds is nowhere near enough for software to be trustworthy. It's why even Orange Book had more than one sentence in its feature and assurance activities recommendations.
Let's put it to the test though. If Im right, most the majority of OSS software will be similarly to proprietary full of easily-prevented holes, undocumented/barely-clear functionality, and difficulty even building it. Whereas high assurance systems would've had the opposite attributes while faring well during professional pentesting w/ source.
One of us was right for a decade straight. Maybe it's because the principles and practices I promoted... work? Evidence in the field is on my side. Neither OSS nor good builds are enough.
There's more than process going on here; there's also a policy axis in this graph.
Closed source has no obligation to reveal vulnerabilities, fix anything, or even work with customers who report vulnerabilities. ORCL will sue you [1] if you learn too much about what you bought. It's often more in their interest to fix something after public discovery for PR reasons but leave it on the todo list otherwise.
So yes, of course closed and open source have holes. The question is, will they be found, announced, and addressed or will they lay secret for years behind a legal wall?
Thats a good point. I should factor that into the next version of my essay. It will be a natural advantage for OSS. On proprietary end, perhaps the contract will give disclosure of problems plus immunity to suits to the reviewer. There are still problems that can happen in that model but negates most of what you said.
Btw, about obligation, my essay assumes the company is trying to differentiate by taking initiative and having their product reviewed. Companies that don't shouldn't be trusted at all. End of story.
In that case, of course you get into mitigation options.
Depending on your contract, proprietary vendors offer few choices about getting a vulnerability patched, if ever. If you're in the riffraff section (ie, most router owners only have a few), you might wait a very long time. One [1] from netgear that languished for months. And what about the Juniper backdoor: won't fix?
With open source, you can take the code to whomever you wish, fix it in house, offer a bounty, etc etc. There are plenty of houses that give away code and sell support. If GPL'ed, this model also accelerates fixes because everyone gets immediate benefit of everyone else's fixes.
Hmm. I'm probably going to have to rewrite or supplement the essay to account for these other dimensions. So, what do you think of it's primary intent to show that belief in security or no subversion come down to trusting a reviewer and methods put in rather than source access? In general rather than for, say, a project whose source you personally would review in full for every kind of security issue, etc.
You're not wrong that a correct process dramatically limits classes of issues (I've worked in a very high ceremony requirements-tracability shop). But it still requires the customer to have faith in source code they can not themselves see.
As a customer, you can claim what you like about your process and your glory: until I can actually verify the code as part of the deliverable, it's just faith on my part. I'd rather use an open source firewall rather than a closed source firewall, regardless of the claims made by the proprietary company. Again, this is about faith. I'd like to avoid having it when it comes to security.
"But it still requires the customer to have faith in source code they can not themselves see." (pnathan)
"4. Low-level, simple, modular code mapping to that.
5. Source-to-object code verification or ability to generate from source on-site." (me)
Seriously, did someone hack my comment where it doesn't show that on everyone else's end or did they hack my system where 4 and 5 are only visible to me? Shit! Here I was using OSS, reviewed, well-maintained software specifically to reduce the odds of that. I'm blaming Arclisp: must have called a C function or something.
"You're not wrong that a correct process dramatically limits classes of issues (I've worked in a very high ceremony requirements-tracability shop)."
Well, there we go. Least you saw that and have experienced that assurance activities can increase assurance. Now we're getting somewhere.
"Again, this is about faith. I'd like to avoid having it when it comes to security."
You're probably going to have it anyway unless you specifically verified the software, libraries, compiler, linker, build system, and all while producing it from a compiler you wrote from scratch. Nonetheless, open-source can increase trust but I say closed can be more trustworthy. Not is or even on average but can be.
Here's my essay that claims and supports that the real factors that are important are the review, the trustworthiness of the reviewers, and verification you're using what they reviewed. I'd like your thoughts on it as I see where you're coming from and like the faith angle. Faith in the integrity of the process and reviewers are the two things I identified as core to security assurance. So, I broke it down to give us a start on that.
I don't think most understand what (4) and (5) are, which is why you're seeing the responses you're seeing. Not being in that field, I had to re-read it a couple of times to understand what they meant.
I think OpenSSL's past disproves many of the pro-OSS claims.
As most things, a blended approach is probably best. Defense in depth, layers of security, crunchy on the outside, still tough on the inside. If you put all your eggs in one basket, you're gonna have a bad time, unless it was a very expensive, well-engineered basket.
That could be the problem. If it was, I take by what was in the comments to those people. Will have to make it more clear next time. No 4 was source code that maps directly to specs, high-level design, requirements, whatever. Modular, code that clearly belongs. No 5 either means generating the system onsite from source-code or an audit trail that goes from source statements to object/assembler code so you can see they match.
"I think OpenSSL's past disproves many of the pro-OSS claims."
100% agreement.
"Defense in depth, layers of security, crunchy on the outside, still tough on the inside. If you put all your eggs in one basket, you're gonna have a bad time, unless it was a very expensive, well-engineered basket.
And even then you might still have a bad time."
Decent points. Other engineers and I went back and forth on discussions involving the latter point due to all the factors involved. A high assurance design usually worked pretty well. Yet, it might not, so Clive Robinson and I's consensus was combining triple, modular redundancy w/ voters and diverse implementation concepts. So, three, different implementations of concepts that shouldn't share flaws with at least one high assurance (preferably three). The voting logic is simple enough it can nearly be perfected. Distributed voters exist, though.
Shit gets worse when you think of the subversion potential at EDA, mask-making, fab, and packaging companies. You have to come up with a way to trust (or not) one entity. For HW, esp low hundreds of MHz, the diverse redundancy with voters should do the trick. Until the adversary breaks them all or the voter. Lol...
Not really. I mean, open source code is definitely preferable in addition to any other security guarantees, for multiple reasons. But an open source spec of a closed source implementation, together with a machine checkable proof (that works on the binary itself), would be equally if not more valuable as assurance that the system performs as intended.
INFOSEC started with Ware Report defining issues and predicting requirements. Lots of CompSci and experimentation by military followed. MULTICS pentest and other landmark works happened. Most key activities that boost security were identified then included into standards like Orange Book. A niche market and CompSci field flourished studying and building systems that used every technique known to improve security with great results (on security anyway). U.S. government canceled mandate on high assurance to switch to COTS for features & cost. Private market did same for similar reasons. High assurance market imploded with research going to a trickle.
PC and Internet eras were really getting going around that same time. Languages and approaches that introduce vulnerabilities one way or another became huge. INFOSEC research and products shifted toward all kinds of tactics for hardening, analyzing, monitoring, and recovering such inherently, insecure stuff. Revisionism kicked in where people forgot the old wisdom, starting to reinvent it slowly, plus how and why they got there in the first place. The products, both regular and security, had tons of vulnerabilities old methods and tools prevented. I call this the Modern Era of INFOSEC. It's still running strong.
Good news is the Old Guard published tons of papers and experience reports telling us what to do, what not to do, and so on. There's a steady stream of CompSci people and some in industry building on that. Keeps advancing. Even mainstream IT and INFOSEC adopted some of the strategies. Rust, "side channels" analysis, unikernels, trusted boot... all these are modern variations (sometimes improvements) on what was done in 70's-early 90's. So, it's not dead but it's mostly ignored and barely moving.
That's what I'm thinking when I see another modern firewall or whatever with less security than the guards from the 80's that predated them. You'd think they'd have learned something by now past just the features. The assurance activities were there for a reason.
Guards, if you were wondering...
https://en.wikipedia.org/wiki/Guard_%28information_security%...
Good essay on security assurance from engineering rather than subjective point of view that development often takes:
We develop a tool to verify Linux netfilter/iptables firewalls
rulesets. Then, we verify the verification tool itself.
Warning: involves math!
This talk is also an introduction to interactive theorem
proving and programming in Isabelle/HOL. We strongly suggest
that audience members have some familiarity with functional
programming. A strong mathematical background is NOT required.
TL;DR: Math is cool again, we now have the tools for "executable
math". Also: iptables!
does it? so you can say "yes, we do not allow these ports" or whatever... does that mean it is "secure"? this doesn't check iptables, only the rules that you write using iptables (as far as I can tell)
the point I am getting at is that in a firewall product, what you are checking is something like "only those users listed from this source may login" which is "easy" but other more complicated things like "this channel must be encrypted". what does "encrypted" mean? Is it just a word that you use in your specification language? if so, does it mean what you think it means? etc etc etc...
formally proving stuff about the behavior of a system, or a distributed system, is hard, but formally proving stuff about security, especially as it relates to information flows, side channels, etc is veryhard...
""yes, we do not allow these ports" or whatever... does that mean it is "secure"?"
No, it means you don't allow those ports. That's all it means. Saying precisely what you're doing or what attributes your aiming for are what the formal work is all about. Whether your security policy is enough and your design embodies it are different things altogether.
"what does "encrypted" mean? Is it just a word that you use in your specification language? if so, does it mean what you think it means? etc etc etc..."
That's actually one of the easiest things you could ever check in formal systems. It's helpful to think of it like programming. Actually, done side-by-side. You implement a formal spec for a requirement and/or high-level function that takes plaintext as input and outputs ciphertext. This typically uses Red/Black separation model. You also produce and vet an implementation for the encryption module. Now, how to know if something was encrypted in the system before going on the wire? Wait for it... wait for it... answer: it went through the encryption module first while it was initialized and in a state saying it's encrypting. Just like that. Labels and checking at interfaces were used for identifying what went with what and making policy enforcement easier.
Note: Guttman has a security kernel in cryptlib that does this at the interface level for every function in the system. In CompSci literature, it's called Assured Pipelines if you want to look it up. Easier to support now with Design-by-Contract and advanced typing systems. Past systems were kludgy when they happened at OS level except with capability systems.
Anyway, such components and functions are composed with the interactions and compositions analyzed. Each component and composition usually has a small number of execution traces it can perform so one can brute force the analysis if necessary. Finite state machines, both success and fail states, were common in high assurance development because they can be analyzed in full in all sorts of ways.
Note: It was Dijkstra that invented the method for this in the THE multiprogramming system. I think PSOS did it for security first. VAX Security Kernel Retrospective has nice sections where they show layering and assurance activities plus effect they had on analysis and defect hunting. Google any of those to understand the method better.
> You implement a formal spec for a requirement and/or high-level function that takes plaintext as input and outputs ciphertext.
this misses the relationship between the size of the plaintext and the ciphertext...
edit: part of the problem is tooling, to deal with stuff like this you need dependent types and those do exist, but not in a way that "the programmer on the street" can use...
One of the best in the field. :) Although I think they just have dev assurance. Not total high assurance. An early one with total was GEMSOS by Schell, one of INFOSEC's founders:
A comparable one for dev assurance, but maybe easier to emulate, was CompCert. Testing of it against many other compilers validated formal verification gave best reliability.
That's right: a formal specification, formal security model, and proof the spec implements it. An implementation formally proven to implement that spec will then posses the security property unless done in by stuff not covered by that model.
Which is where EAL6/7's other assurance activities come in.
Amazon notes many kinds of problems that lead to reliability and security failures that TLA+ helped knock out. Their engineers are sold on it now and have no intention of dropping it. Certain sentences in that article are nearly identical to those I read in ancient papers using ancient methods. The benefits of precise, easy-to-analyze specifications are apparently timeless.
Here's it done in Z via Altran's Correct by Construction methodology:
They apply about every lesson learned in high assurance in their work. Defect rate, from this demo to old Mondex CA, is usually around 0.04 per 1,000 lines of code. That's better than the Linux kernel.
Rockwell-Collins formalized a separation architecture, HW, microcode, etc then integrated it into a CPU:
NICTA, who did seL4 verification, use tools to model stuff in the language that causes security errors then use provers to verify they're used correctly. Example tool:
Lots of groups using lots of different tools with great results. The difficulty and impact on time-to-market varies. The use of compositional, functions or state-machine models with subsets of safe languages, design-by-contract (or just interface checks), static analysis, design/code review, testing, and compilation with conservative optimization seem to be the winning combo. There's free tools for all of that. It takes about the same time to code as usual while preventing lots of debugging during testing and integration.
1. Clear description of every feature and requirement in system.
2. Mathematical spec of those where English ambiguity could effect results.
3. High level design with components that map 1-to-1 to those.
4. Low-level, simple, modular code mapping to that.
5. Source-to-object code verification or ability to generate from source on-site.
What people in faux security mocked as mere "paperwork" or "red tape" were actually pre-requisites for defeating subversion my mentally understanding a system from requirements all the way to code. A problem like this would've been impossible in such a system because it would be beyond obvious and probably unjustifiable with requirements tracing.
Every story like this further validates the methods that consistently produced systems without all the security problems plaguing modern security products. Situation isnt inevitable or even necessary: merely an inversion of scientific method where security companies and professionals consistently refuse to use what's proven to help and reuse tactics proven to fail. It's gotta stop.
That it wont is why I favor liability legislation tied to a reasonable baseline of practices. We can use an inexpensive subset of what worked in highly assured systems. 80/20 rule. Baseline would look more like Secure64 or HYDRA firewall than shit like Fortinet and Juniper. Hackers would work for exploits. I know Im dreaming, though, as DOD and NSA just dropped mandate to EAL1 w/ 90 day review for some stuff. (Rolls eyes).