- Dailydave - mm4.emwd.com

while True:
by Dave Aitel 24 Mar '24

24 Mar '24

There seem to be a lot of people who think the problem with cyber security is we aren't paying lawyers enough. This results in the current push for software liabilities, or the need to click accept on cookies before we use every website. It is natural for lawyers to want to feed the next generation of associates, by regurgitating legal koans into their mouths. These vomitous truisms pass for thought leadership when you go high enough into the cyber policy clouds. "We don't know what we don't know!", "You can't manage what you can't measure!" , "We need to be Secure to Market not First to Market!", "Crawl, Walk, Run!" These statements are the opposite of haikus, which when done right are one crystalized emotional moment. This is why I think maybe we should hire more poets to do cyber policy, instead of lawyers. What is "*i carry your heart with me(i carry it in*" other than the first line to an exploit written in a bash script we all forgot existed but our spirits remembered? When LSD-PL said: bn,a bn,a call did that not carry with it an emotive punch? When we hack, do we not reach within our internal well of hate to pull forth a tiny amount of darkness and then send it into the world on tiny flaxen wings? I cannot do a survey on this in any language that matters, but I look into the net and see all the ancient hackers I grew up with still crouched in full armor, their ocher swords smoldering. This week in between cyber policy calls at 0500, I sat for hours, choking on the byzantine syntax of LangChain attempting to wrestle an LLM into submission. I kept thinking, what would Horizon do? What would Shubs do? What would James Kettle do? What would Tiraniddo do? What would Chompie do? What would Skylar do? I told myself: They would *continue*, is what they would do. -dave

1 0

Value of the [leaked] Windows source
by Konrads Klints 06 Mar '24

06 Mar '24

Windows XP and Windows 2003 partial source code is out there on github. With such a rich corpus of known vulnerabilities in those OS'es and source code availability, surely there should be an amazing amount of SAST/semgrep/codeql rules that take as input existing known exploits and then do rules that find similar things, yet I don't seem to be able to find such projects Surely, these two code bases should be the foundation of most good CS/cyber courses - like students finding new bugs, etc. Is source code junk?

2 1

0xC15A: Secure By Design and Secure by Default
by Dave Aitel 23 Jan '24

23 Jan '24

So I have a ton of thoughts on the CISA Secure by Design and Secure by Default push that is ongoing, as I am sure many of you do. And the first thought is: This is not a bad way to go about business as a government agency in general. I think it's easy to ignore how fast the USG has changed its business practices, showing an agility that few large organizations can match. In particular using Secure By Design as a case example. 1. Massive outreach to garner feedback (including at defcon, but also via email, etc.) 2. Multiple rounds of editing of proposals 3. Actual people you could call and talk to about the proposal, with their faces and positions listed right in the papers and blogs and lawfare podcasts. If you were in DC today you could probably hit one of them up for drinks or lunch or whatever. 4. Interaction across multiple stakeholder groups, including internationally 5. The "right people" involved - and you can tell their backgrounds from what they are annoyed about during their podcasts and other presentations. (i.e. Bob Lord is very annoyed about XSS and obsessed with car safety, which I'll dig into later). But also Jack Cable, Lauren Zabierek and Grant Dasher are all worth listening to. 6. Clear executive support So that's all good stuff. I thought I would post it as its own note because it's rare to spend a moment to look at the government process, and not see literally sausage being made. :) -dave

3 3

Secure By Default Part 2
by Dave Aitel 19 Jan '24

19 Jan '24

So I wrote a little draft essay on Secure By Default and opened it for comment. I think one thing that we maybe forget in our community is that some of the more fundamental basises of what we do never make it up to policy-world. Langsec being the primary example. But also there's a huge body of work in TAOSSA, Shellcoders, every offensive conference talk, etc. that never gets put into context anywhere but in our clique. Obviously feel free to just comment in-thread if you prefer, even if you work at CISA: https://mastodon.social/@dave_aitel/111779922142416342 Thanks, Dave

1 0

Leverage
by Dave Aitel 27 Dec '23

27 Dec '23

So we know a lot of people who've gone into Big Corpo or sold a company or just worked hard and gotten lucky and happen to be richer than the average bear. And while a lot of those people put their money into nice things, nothing wrong with that, a lot of them also try to use that money to change the world, and then they find out it's harder to change the world with money than it is with an exploit. And I know a lot of people who say this out of experience. I used to say, and I continue to say, that most cyber policy experts have never seen a real exploit. Yes, even the news reports of 0days are addictive. They have cool names, like a street drug, they have a shadowy underworld, they have a bodyguard of rumors and insinuations. Literally, as I write this, Kaspersky is at CCC doing a street barker presentation on how much someone thought they were a legitimate target worth throwing down an iOS 0day chain on, which frankly is not something I would brag about as a nominally defensive company. Part of the problem is we analyze exploits out of failures. Reverse engineering an exploit does not show you the exploit any more than dissecting a Humboldt squid can show you the terrors of the deep. Once you've seen an exploit change the world, you are forever stunned. It is impossible to go backwards to your previous life. It is like your first taste of sex or love or the feral inhuman joy of combat. For some people, it's probably better. But how exploits really work has shockingly little to do with the circus around exploits on social media or in the news, or press releases from endpoint security companies or government agencies. And the same is true with philanthropy. Here we have "Give Miami Day!" where some random billionaire will match your funds if you give to your school, which should be getting all the government funding it needs, but clearly isn't. I don't know if this is at all useful to be honest. It feels more like covering a hole your government put in your school budget on purpose. I tried buying MEL science kits for entire classes of my local grade school. It worked once, and one of the kids was like "I didn't even know I liked science". But it was largely impossible to get the kits USED by the teachers, who are under levels of logistical stress that would stymie a Marine platoon. So I ended up giving up on this effort. I supported Project Grapple, which worked well because they have a leader at the top who is focused on the success of a small set of kids in a very personal way. But those kinds of leaders are almost impossible to find. When you find them, it's the best investment you can make if you want to change the world one kid at a time, which might be the only way. And these leaders don't last forever. A lot of big donors focus on things like FIRST Robotics, which frankly has been a massive success and offers a lifeline for kids in schools where nothing else matters or meets with any success. It's prohibitively tricky to figure out which schools have a local leader that can take money and build a robotics club out of it. It's very much not important that the kids WIN the competitions - which at the top level are between Northrop Grumman, Raytheon, and Lockheed Martin engineering teams doing ad-hoc group apprenticeships. So again, finding these leaders is a constraining problem. Just like with exploits, what I find is that changing the world with philanthropy is targeted , personal, and more complicated than it looks. I have seen very smart people struggle with finding leverage in this space. Nonprofits are themselves often quite exploitative of their employees or just generally ineffective. Anyways, what I'd like to see is the Gulas and Syversons and Alperovitch's and Pollocks and so forth put a Slack channel together to build a bit of a body of work on how to do this correctly. A bugtraq for changing the world with dollars, if you will. -dave

2 1

A holiday reflection on: Training.
by Dave Aitel 19 Dec '23

19 Dec '23

I think one thing this community does really well, better than almost any other community I've found, is training. It's amazing, in a way, because this is a community of professional secret holders. And yet everywhere you look, a hacker is putting their heart and soul into iterating on lab exercises for their class in whatever sub-field they are an expert in. And giving training is hard. It's hard in the way consulting is hard, but with even more social activity. On one hand: It's lucrative? But hour for hour, you're probably better off financially by finding a new bug or doing consulting work, or any number of other activities than building, marketing, and running a training class. When I left my last position, one of the first things I did was pay for and take Amy Burnette's browser exploitation class. And that's paid off to this day, really. And there's so many good classes, taught by all the specialists in our various sub-niches. It's spectacular that in this world of auto-didacts, we are gifted in the quantity and quality of training available in our field in a way that is basically unheard of in any other field. Of course, there's a lot of things you can't learn from training, and I was reflecting on this while sitting down and reading the labyrinthine specifications of some huge protocol for one of my current projects. A lot of the best bugs I've ever seen hackers find have been from doing exactly that: They sit and hit page down on some extensively huge and boring documentation with the steady, persistent rhythm of a neurodivergent woodpecker pecking at a tree, each tap bringing them closer to the elusive kernel of truth. Like, I know people who have various protocol RFCs printed out for long airplane rides. I've seen hackers read through a book on an operating system design and then just circle an LPE in the book with a yellow highlighter. On the flip side, there are times when you dive headfirst into a colossal specification, emerging as a veritable guru on an esoteric legacy mail transfer mess like X.400. Yet, despite this newfound expertise, you find yourself no more enlightened or advantaged than before, as if you've scaled a mountain only to find the summit shrouded in the same thick fog that cloaked its base. Anyways, happy holidays everyone. Hopefully you had a year of worthy discoveries. -dave

1 0

t2'24: Call For Papers 2024 (Helsinki, Finland)
by Tomi Tuominen 17 Dec '23

17 Dec '23

Call for Papers 2024 t2 infosec has been pushing the boundaries of security research for two decades and it don't stop. We're back April 18-19, 2024 - Helsinki, Finland. CFP and registration are both open. This is an event for newcomers, established merchants of dual use computer code, beginners of vulndev, cyber vagabonds, retired or redacted, and hackers of all sorts. If you have new original security research targeting old, current or future technology, please submit: https://if.t2.fi/action/call_for_papers Got 99 problems but an 0day ain't one. Helsinki offers a Northern European mood, with the resilience built by using Linux on the desktop since 1991, fueled by IRC, demoscene, VHS tapes, invention of modern backpropagation, a lot of memes and sauna. This is the country, which exports Alan Wake 2, quantum computers, and solitude. As regularly as seasons change, new technology is introduced, vulnerabilities - old and new are discovered, lessons of tradecraft are learned and teachings of the old slowly forgotten.To foster the passion for advancement of (in)security and to keep that fire burning bright for future generations, we gather an intimate crowd of people every year to enjoy the advance of offensive research, share lessons of coming back from the edge, and build those valuable human connections. Whether your research tackles LLMs, qubits, Azure/AWS/GCP, iOS internals, the market leading or lagging EDR products, access control systems, SoC, BTLE, tipping ChatGPT or something else, which is surprising to either humans or computers, we'd like to hear about it. Security researchers, you know the rules and so do we. This is an event for the community, by the community. Our focus is on technical excellence, not politics or player hating. Come as you are. The advisory board will be reviewing submissions until 2024-02-04. Slide deck submission final deadline 2024-04-02 for accepted talks. First come, first served. Submissions will not be returned. Quick facts for speakers + presentation length 60-120 minutes, in English + complimentary travel and accommodation for one person[6] + decent speaker hospitality benefits + no marketing or product propaganda Still not sure if this is for you? Check out the blast from the past. !!LINK!! [0] hunter2 [6] except literally @nudehaberdasher and @0xcharlie Call for Paper submissions https://if.t2.fi/action/call_for_papers Buy a ticket to the conference https://if.t2.fi/action/register/attendees -- Tomi 'T' Tuominen | Founder @ t2 infosec conference | https://t2.fi

1 0

what is reasoning?
by Dave Aitel 05 Oct '23

05 Oct '23

There's not a huge paper I can find on the history of all the big bug classes. I mean, there's probably not even a real definition of "bug class" that would survive a drunken conversation between two CTF teams. But like, screw it, "format strings" are a bug class. "Deserialization bugs" are a bug class. "strcpy and friends" is a bug class. You know what I mean if you're the type of person who subscribed to this list. One thing you know if you internally describe yourself as "hacker" and not "security professional" is that the bug classes of the future are here now, but the commercial world doesn't see them yet or sees them as one-off issues. For example, check out this amazing slide from this GoSecure presentation ( https://gosecure.github.io/presentations/2019-04-29_atlseccon/History_of_De… ) [image: image.png] If you can't see the picture, it was 9 years from "first public instances" to "bell curve of public mass use of bug class". The nightmare for the commercial industry of course is that this ignores private research, which looks like a much bigger bell curve, superimposed on the whole thing as you all know. If you've been working on writing automated tools to find bugs, then you also know the hardest thing to find is a new bug class. We can't all be Rain Forest Puppy! But even building things to replicate how hackers think when they just want to find Deserialization or strcpy bugs is hard. Like, what does it mean to even understand a bug? Bugs are not on one line! They are not even one concept! Can you inject a bug into a C program without also injecting another bug you didn't even intend on injecting? Even though humans are terrible at logic, we somehow think that stepping logically through a problem is how reasoning works. But you have to get over this! Stare at an ant farm and chant to yourself the neo-Buddhist mantra of "This is reasoning. This is cognition" and this way you quickly realize that precision is not the answer when you try to go from building a calculating machine to a reasoning machine. Precision and reasoning are opposites! In other words, the closest we have to reasoning over program behavior is fuzzing. Once you understand that, a lot of other stuff starts to make sense. -dave

1 0

DARPA AI Challenge!
by Dave Aitel 29 Aug '23

29 Aug '23

I've been working with LLMs for a bit, and also looking at the DARPA Cyber AI Challenge <https://www.darpa.mil/news-events/2023-08-09>. And to that end I put together CORVIDCALL which uses various LLMs to essentially 100% find-and-patch any bug example I can throw at it from the various GitHub repos that store these things (see below). [image: image.png] So I learned a lot of things doing this, and one article that came out recently (https://www.semianalysis.com/p/google-gemini-eats-the-world-gemini) talked about the future of LLMs, and if you're doing this challenge you really are building for future LLMs and not the ones available right now. One thing they pointed out in that article (which I highly recommend reading) is that huggingface is basically doing a disservice with their leaderboard - but the truth is more complicated. It's nice to know which models do better than other models, but the comparison between them is not a simple number any more than the comparison between people is a simple number. There's no useful IQ score for models or for people. For example, one of the hardest things to measure is how well a model can handle interleaved and recursive problems. If you have an SQL Query inside your Python code being sent to a server, does it notice errors in that code or do they fly under the radar as "just a string". Can the LLM handle optimization problems, indicating it understands performance implications of a system? Can the LLM handle LARGER problems. People are obsessed with context window sizes but what you find is a huge degradation of accuracy in following instructions when you hit even 1/8th the context window size for any of the leading models. This means you have to know how to compress up your tasks to fit basically into a teacup. And for smaller models, this degradation is even more severe. People in the graph database world are obsessed with getting "Knowledge graphs" out of unstructured data + a graph database. I think "Knowledge graphs" are pretty useless, but what is not useless is connecting unstructured data by topic in your graph database, and using that to make larger community detection-based decisions. And the easiest way to do this is to pass your data into an LLM and ask it to generate the topics for you, typically in the form of a Twitter hashtag. Code is unstructured data. If you want to measure your LLM you can do some fun things. Asking a good LLM for 5 twitter hashtags in comma separated value format will work MOST of the time. But the smaller and worse the LLM, the more likely it is to go off the rails and fail to do it when faced with larger data, or more complicated data, or data in a different language which it first has to translate. To be fair, most of them will fail to do the right number of hashtags. You can try this yourself on various models which otherwise are at the top of a leaderboard, within "striking distance" on the benchmarks against Bard, Claude, or GPT-4. (#theyarenowhereclose, #lol) Obviously the more neurons you have making sure you don't say naughty things, the worse you are at doing anything useful, and you can see that in the difference between StableBeluga and LLAMA2-chat, for example, with these simple manual evaluations. And this matters a lot when you need your LLM to output structured data <https://twitter.com/RLanceMartin/status/1696231512029777995?s=20> based on your input. So we can divide up the problem of automating finding and patching bugs in source code in a lot of ways, but one way is to notice the process real auditors take, and just replicate this by passing in data flow diagrams and other various summaries into the models. Right now hundreds of academics are "inventing" new ways to use LLMs. For example "Reason and Act <https://blog.research.google/2022/11/react-synergizing-reasoning-and-acting…>". I've never seen so much hilarity as people put obvious computing patterns into papers and try to invent some terminology to hang their career on. And of course when it comes to a real codebase, say, libjpeg, or a real web app, following the data through a system is important. Understanding code flaws is important. But also building test triggers and doing debugging is important to test your assumptions. And coalescing this information in, for example, the big graph database that is your head is how you make it all pay off. But what you want with bug finding is not to mechanistically re-invent source-sink static analysis with LLMs. You want intuition. You want flashes of insight. It's a hard and fun problem at the bigger end of the scale. We may have to give our bug finding systems the machine equivalent of serotonin. :) [image: image.png] -dave

1 0

Book Review: Fuzzing Against the Machine (Antonio Nappa, Eduardo Blazquez)
by Dave Aitel 23 Aug '23

23 Aug '23

https://www.packtpub.com/product/fuzzing-against-the-machine/9781804614976 The authors claim in their conclusion: "We want to stress the importance of books as journeys to explore and experience topics from the unique viewpoint of the authors." And in this they succeeded. This book works best as a proposed curriculum for a five day workshop for experts to reproduce fuzzing frameworks that target embedded platforms - including Android and iOS. Largely this is done by figuring out how to get various emulation frameworks (QEMU in particular) to carry the weight of virtualizing a platform and getting snapshots out of it and pushing data into it. Fuzzing is a childishly easy concept that is composed of devilishly hard problems in practice (7 and 8 being the ones this book covers in depth - the fuzzers themselves are simplistic other than those topics): 1. Managing scale 2. Getting decent per-iteration performance 3. Triaging crashes 4. Building useful harnesses 5. Knowing when you have fuzzed enough, vs. being in a local minima 6. Figuring out root causes 7. *Getting your fuzzer to properly instrument your target so you can have coverage-guided fuzzing* 8. *Handling weird architectures* 9. Generating useful starting points for your fuzzer (or input grammars) All of these things are basically impossible in the real world. Your typical experience with a new fuzzing framework is that you install it on a fresh Linux, pick a target, and then watch as it fails to instrument or even run. In other words, just knowing which fuzzer versions to use, and on what, is valuable information. When I read a book on security, a good one, I want it to feel like I'm putting on a brand new powersuit, ready to march into the wilderness with a flamethrower and a mindset of extreme violence. This book delivers that feeling. Because while my current business practices have nothing to do with fuzzing the Shannon baseband, that doesn't mean some small part of me doesn't want to. We all have the dark urge. We crave SIGSEGV in things people rely on. So in summary: 10/10, great book. Would recommend buying 10, setting up a class, and going over it all together. Of course, this field is RAPIDLY EVOLVING and you're going to want to get it updated, perhaps with the fancy new PCODE fuzzer Airbus released earlier today. ( https://github.com/airbus-cyber/ghidralligator) -dave

1 0

2025

2024

2023

2022

2021

2020

Dailydave