主题演讲:剖析全球最流行容器所学到的经验 - Ayse Kaya
本页标题、字幕、摘要和阅读笔记为已保存的 AI 译文,可能有误。时间戳与来源一致,可展开查看原始字幕,并对照原视频核对。
Ayse Kaya 审视容器扫描的发现、漏洞趋势和软件复杂性,并追问安全实践如何支持开发者。
转录来源:Vidleaf上的自动语音识别(可能包含错误)。AI标点由Vidleaf添加;保留原始措辞和时间戳锚点。原始转录可在下方获取。
AI生成的摘要
初创公司 Slim.ai 的一位数据科学家审视容器扫描数据揭示了开发者所体验的容器现状。Slim 今年已扫描了超过 900,000 个容器,发现显示 60% 的顶级公开可用容器比一年前有更多漏洞,高危漏洞上升了 50%,而如今平均每个容器有近 300 个漏洞,其中约 30% 个为高危或严重,高于去年的 20%。组件复杂性也在增长:平均每个容器有 400 个软件包,许可证增加了 2.5 倍,元数据上升了 40%。在一项全球随机研究中,70% 的开发者和 DevOps 工程师表示客户正在要求或强求零漏洞的容器,88% 表示移除漏洞正变得越来越具有挑战性,复杂性是首要因素;只有四分之一的开发者表示他们完全理解容器瘦身和加固,而 49% 的高管表示其公司正在对容器进行瘦身和加固,而实际从事这项工作的人报告的数字则显著更低。演讲者认为复杂性不是敌人,无知才是,并表示 Slim 旨在通过自动化和智能优化为全世界所有容器解决这一问题,并表达乐观,因为世界各地有坚持不懈、能力出众的团队正在研究这些问题。
根据文字记录生成。请对照原始视频检查链接段落。
带时间戳的章节摘要
复杂性、容器与 Kubernetes
- [0:12] 演讲者从小就对复杂性着迷,尤其是它如何从简单的开端中涌现。
- [0:28] 作为一位专注于新兴技术的 MIT 研究员,她被 Stephen Wolfram 和 John Conway 的《生命游戏》等研究深深吸引,这些研究将计算机系统与生物世界乃至生命本身联系起来。
- [0:56] 她认为 Kubernetes 可能比生物系统更复杂。
- [1:02] 如果 Kubernetes 是生物体,那么容器就是细胞,不过她承认这个类比可能有些过头。
- [1:14] 容器无时无刻不在被使用,无处不在,几乎遍及所有垂直领域的每一个软件开发项目。
- [1:24] 她不是 DevOps 工程师,而是一位为名为 Slim.ai 的初创公司工作的数据科学家。
Slim 的容器扫描数据
容器漏洞趋势与开发者压力
容器、开发者与 Docker 的采用
- [3:44] 一个故事正在展开:容器正变得更容易受攻击,客户比以往任何时候都更要求完美,而开发者感受到了压力。
- [4:05] 开发者对 Kubernetes 和容器的兴趣,以及大大小小公司对容器的采用,正不可否认地高涨。
- [4:22] 开发者开始将 Docker 视为其工作流程的原生组成部分,就像 Git 或 IDE 一样。
- [4:31] 根据 Stack Overflow 的 2022 开发者调查,Docker 是开发者最喜欢且最想学习的头号技术。
- [4:46] Docker Hub 上的历史总拉取量在短短一年内增长了两倍。
- [4:52] 虽然许多教程涵盖容器的交付和构建,但人们对容器内部的东西了解甚少。
- [5:07] 容器被视为原子单元和生物体的构建块,请求进入,数据出来。
Log4shell 与容器安全意识
Docker Hub 顶级容器的安全发现
容器漏洞与复杂性趋势
- [7:23] 一份后续报告重点关注去年与今年之间的差异。
- [7:36] 第一个发现:所有类别中的高危和严重漏洞比以往任何时候都多。
- [7:45] 60%的容器如今比一年前有更多漏洞。
- [7:58] 新事件的检测速度是修复速度的四倍,而且新事件大多是高危和严重级别。
- [8:11] 今年高危漏洞增加了50%。
- [8:20] 平均每个容器有近300个漏洞,其中30%是高危或严重级别,高于去年的20%。
- [8:43] 第二个发现:组件复杂性显著增加;平均每个容器有400个软件包,而这只是冰山一角。
- [9:14] 许可证增加了2.5倍,容器有更多层且更大,扫描时间更长,元数据增加了40%。
- [9:38] 这些工具和软件包对于实验是必要的,有助于开发者调试、测试、构建和娱乐,但它们也代表了攻击面,并意味着需要一个自动化流程来移除它们。
容器安全与工程师差距
容器安全与未来乐观主义
[0:00] 你是否曾感到无限渺小。在某种广阔而复杂的事物面前。就像……午夜的天空。从小女孩时起,我就对复杂性着迷。尤其是复杂性如何从简单、卑微的起点涌现。从基本粒子到星系。作为麻省理工学院专注于新兴技术的研究员,我着迷于。
[0:34] 被斯蒂芬·沃尔夫勒姆和约翰·康威的生命游戏等研究所吸引。将计算机系统与生物世界联系起来。与生命本身联系起来。毕竟,还有什么能比生物系统更复杂呢?对吧?也许除了Kubernetes。如果Kubernetes是生物体,那么容器就是细胞。也许这个类比有点太过了,但我们知道容器一直在被使用,无处不在。
[1:17] 在几乎每个软件开发项目中,跨越所有垂直领域。看?我不是DevOps工程师。我是一名数据科学家。为一家名为Slim.ai的初创公司工作。在Slim,我们痴迷于容器优化和安全。我们定期扫描数十万个容器。仅今年一年,我们就扫描了超过900,000个容器。
[1:50] 作为一名数据科学家,处理这些数据,我想为你提供一个了解容器世界的窗口。正如开发者所经历的那样,通常是第一次。我想让我们扩展对这个世界的感知。我想让我们看一些关于这个世界的视角转换数据。以60%为例。
[2:23] 那就是。如今比一年前拥有更多漏洞的顶级公开可用容器的百分比。60%。这是在经过一年对软件供应链安全的强烈关注之后。在多次安全事件之后。谢谢。其余的容器并没有改善那么多,我们稍后会讨论这一点。
[2:56] 但在此之前,让我们将这个见解与这里的另一个统计数据结合起来。这个。来自我们设计的一项全球随机研究。向开发者和DevOps工程师提问。70%表示他们的客户正在要求、要求他们的容器零漏洞。除此之外,88%的开发者表示从容器中移除漏洞变得越来越具有挑战性。
[3:34] 最主要的因素是复杂性。众多组件及其依赖关系。它们。所以如你所见,一个故事正在我们眼前展开。容器正变得越来越脆弱。客户比以往任何时候都要求完美。而开发者们正感受到压力。尽管开发者们身处战斗的热潮中,但有一股不可否认的上升趋势。
[4:08] 当谈到开发者在Kubernetes和容器方面的兴趣,以及各种规模公司对容器的采用时。事实上,开发者们开始将Docker这样的技术视为其工作流程的原生工具,就像Git或IDE这样的工具一样。根据Stack Overflow的2022开发者调查,“Dakar”是开发者最喜爱的头号技术?
[4:42] 并且想要学习。你也可以在使用统计数据中看到这一点。Docker Hub上的总拉取量在短短一年内增长了两倍。但是,尽管有很多教程可以学习如何交付和构建容器,人们对容器内部的内容却知之甚少。就像自我一样。容器被视为这些原子单位。是活生物体的构建块。
[5:15] 请求进入的地方。数据出来。就这么简单。但然后……像log4shell这样的事情发生了。在我们意识到之前,2022成为了软件供应链安全之年。我们都更加投入于理解我们软件系统的主要组件,以及它们的二阶效应。
[5:48] 它们的依赖树。那么,这种重新觉醒的意识如何改变了容器格局。我的研究团队。是Slim.ai。一直在解构、分析、观察容器,以理解是什么让这些容器对开发者友好且生产就绪。去年大约这个时候,我们将放大镜对准了Docker Hub上最公开可用的容器。
[6:23] 并发布了带有分析见解的报告。我们在那份报告中有些令人震惊的发现,我不打算全部列出。但我要告诉你这一点。作为容器爱好者,我们预期会看到一些异常值。某些容器、某些容器类别,拥有大量漏洞、包、库、许可证、特殊权限等等。但即使是平均值也高得惊人。
[6:57] 关于这些容器需要注意的一点是,是的,Docker Hub上有10百万个镜像,拉取量超过318十亿次。但前165个容器占了所有使用量的30%以上,所以这是一个非常特殊的集群。所以今天,发布了那份报告的续集。这次,聚焦于去年和今年之间的差异。
[7:31] 那么让我们继续查看数据和我们的一些学习发现。第一个发现是,我们在所有类别中拥有比以往更多的高危和严重漏洞。现在,我已经提到60%的容器今天比一年前有更多漏洞。是的,我们确实修复了某些漏洞、某些事件。但我们检测新事件的速度是四倍。
[8:03] 比我们的修复速度。而我们检测到的新事件,新事件,大多是高危和严重漏洞,但高危漏洞今年增加了50%。所以今天,平均容器有近300个漏洞,30%的Fitch是高危或严重。高于去年的20%。
[8:35] 而V认为,20%太高了。让我们进入下一个发现。组件复杂性也显著增加。今天的平均容器有400个包。如果你想一想,“这个数字应该是冰山一角。每个包可能有数千个,有时数十万个依赖项,正如多项研究所显示的。”
[9:06] 但事实证明,即使是冰山一角也是一项艰巨的工作。不仅仅是包。许可证增加了2.5倍,容器有更多层,它们更大,扫描这些容器需要更长时间,甚至元数据。它们是S-bombs。增加了40%。现在,别误会我的意思。
[9:38] 这些工具、这些包对于实验是必要的。它帮助开发者调试、测试、构建和娱乐。但它们也代表了一个攻击面。并暗示需要有一个自动化流程来移除此类内容,以便该攻击面永远不会进入生产环境。数据就是数据。但对我们人员的影响是什么?
[10:10] 这个更复杂且更脆弱的格局如何影响我们的DevOps工程师和开发者。我提到88%的人表示他们发现移除漏洞更具挑战性。只有四分之一。的开发者表示他们完全理解容器瘦身和加固的工作原理。有趣的是,高管和一线工程师之间似乎存在不匹配、脱节。
[10:47] 49%的高管表示他们正在公司中进行容器瘦身和加固。但实际做工作的人报告的數字显著较低。政府和公司可能要求一个零漏洞的世界。但那片土地感觉遥不可及。
[11:18] 鉴于我们的工具和技术。令人清醒的底线是,我们今天并不比2021时更安全。然而,一线希望是,我们已经觉醒。我们比以往任何时候都更意识到这些挑战。对于Slim的我们来说,我们相信你应该知道容器里面有什么。你应该只将所需的内容交付到生产环境。我们正在通过自动化和智能优化为全球所有容器寻求解决这一挑战。
[11:54] 看吧。我并不认为复杂性在这里是敌人。无知才是。有时候我们需要退后一步,欣赏周围复杂系统富有诗意的美。也有时候,我们需要无畏地介入,进行解构和重新设计。我知道。
[12:26] 我们必须解决决策智能的问题。解决这些问题的意志力。而大多数决策者就在我面前。在这个房间里。我的朋友们说我是个无可救药的乐观主义者。但这并不是我认为未来光明的原因。我知道世界各地有不懈努力、才华横溢、能力出众的团队正为这些问题彻夜难眠。
[12:58] 如此多的团队,我甚至无法把仅这个项目所借助的团队名字全部放进我的感谢卡里。而作为人类,当我们聚集在一起,就像我们在这个绝佳的地点所做的那样,当我们把心血和智慧汇聚在一起时,我们甚至能完成极其艰难的任务。这就是为什么我知道,下次我站在这个舞台上时,我会带来。
[13:30] 更好的消息。谢谢。♪ ♪
查看英文原始字幕
[0:00] Have you ever felt infinitely small. in front of something, vast and complicated. Like... the midnight sky. Ever since I was a little girl, I've been fascinated by complexity. especially how complexity emerges from simple, humble beginnings. from elemental particles to galaxies. And as a researcher at MIT focusing on emerging technologies, I was fascinated.
[0:34] captivated by research like Stephen Wolfram and John Conway's Game of Life. Tying. The computer systems fit biological world. with life itself. After all, what could be more complex? then biological systems. Right? Except maybe Kubernetes. If Kubernetes is the organism, then containers are the cells. And maybe that takes the analogy a little too far, but we know that containers are being used all the time, everywhere.
[1:17] In nearly every software development project, across all verticals. See? I'm not a DevOps engineer. I'm a data scientist. working for a startup called Slim.ai. At Slim, we are obsessed with container optimization and security. And we scan hundreds of thousands of containers regularly. This year alone, we have scanned more than 900,000 containers.
[1:50] And as a data scientist, Working with this data, I would like to provide you a window into The world of containers. As developers experience it, often for the first time. I would like us to expand our senses about this world. I would like us to look at some perspective shifting data about this world. Take, for example, The 60%.
[2:23] That's. the percentage of top publicly available containers that have more vulnerabilities in them today than they did a year ago. 60%. And this is after a year of intense focus on social air supply chain security. in the aftermath of multiple security incidents. Thank you. The rest of the containers haven't improved that much, and we'll be talking about that in a moment.
[2:56] But before, let's couple this insight with this other statistic here. This one. is from a global randomized study that we designed. asking questions to developers and DevOps engineers. 70% said Their customers are asking, demanding their containers to have Zero vulnerability. On top of that, 88% of the developers said It is getting more challenging to remove the vulnerabilities from their containers.
[3:34] The number one contributing factor being the complexity. numerous components with dependencies. them. So as you can see, there's a story that's unfolding before our eyes. Containers are becoming more vulnerable. Customers are demanding perfection more than ever. and developers are feeling the pressure. And although developers are in the heat of the battle, There is an undeniable rising tide.
[4:08] When it comes to Developer interests. in Kubernetes and containers. and container adoption across companies large and small. In fact, Developers are beginning to look at a technology like Docker as native to their workflows like a tool like Git or an IDE. According to Stack Overflow's 2022 Developer Survey, "Dakar" is the number one technology developers Love?
[4:42] and want to learn. And you can see this in the usage stats as well. The all-time pools on Docker Hub has tripled in just a year. But while there are lots of tutorials out there, to learn about shipping and building containers. There is precious little understanding around what's in them. Like self. Containers are being seen as these Atomic units. building blocks of living organisms.
[5:15] where requests go in. Data comes out. It's just that simple. But then... Something like log4shell happens. And before we know it, 2022, becomes the year of software supply chain security. And we are all more invested in understanding not only the principal components of our software systems, but also their second order effects.
[5:48] their dependency trees. So how did this renewed sense of awareness change the container landscape. My research team. It's Slim.ai. have been deconstructing, analyzing, observing containers in an effort to understand What makes these containers developer friendly, and production ready. Last year, around this time, We've put a magnifying lens onto the top publicly available containers on Docker Hub.
[6:23] and publish the report with analytical insights. We had some shocking findings in that report, and I'm not going to list them all. But I'm going to tell you this. As container enthusiasts, We were expecting to see some outliers. Certain containers, certain container categories, having a large number of vulnerabilities, packages, libraries, licenses, special permissions, what have you. But even the averages were surprisingly high.
[6:57] And one thing to note about these containers is that, yes, there are 10 million images on Docker Hub, with more than 318 billion pools. But this top 165 containers account for more than 30% of all use, so it's a very special cluster. So today, be published A sequel to that report. This time, focusing on the delta between last year and this year.
[7:31] So let's go ahead and look into the data and some of our learnings findings. The first finding is that we have more high and critical vulnerabilities than ever across all categories. Now, I already mentioned that 60% of the containers have more vulnerabilities today. than they did a year ago. And yes, we did remediate certain vulnerability, certain incidents. But we are detecting new incidents four times faster.
[8:03] than our remediation rate. And the new ones, the new incidents that we detect, are mostly high and critical. vulnerabilities, but the high severity vulnerabilities have increased by 50%. this year. So today, The average container has almost 300 vulnerabilities, 30% of Fitch is high or critical. up from 20% last year.
[8:35] And V thought, 20% was too high. Let's move on to the next finding. Component complexity has also increased significantly. The average container today has 400 packages. And if you think about it, "This number is supposed to be The tip of the iceberg. Every package might have thousands, sometimes hundreds of thousands of dependencies as shown by multiple studies.
[9:06] But it turns out that even the tip of the iceberg is a nice work. And it is not just the packages. The licenses have increased by 2.5x, the containers have more layers, they are larger, the scanning of these containers takes longer times and even the metadata. They're S-bombs. have increased by 40%. Now, don't get me wrong.
[9:38] These tools, these packages are necessary for experimentation. It helps developers to debug and test and build and have fun. but they also represent a type surface. and imply the need to have an automated process to remove this so that that attack suffers surface never makes it production. Data is data. But what's the impact on our people?
[10:10] How does this more complex and more vulnerable. landscape impact our DevOps engineers and developers. I mentioned that 88% said they find it more challenging to remove the vulnerabilities. Only one. in four developers. said they fully understand how container slimming and hardening works. And interestingly, There seems to be a mismatch, a disconnect between executives and the frontline engineers.
[10:47] 49% of the executives said they are slimming and hardening containers in their companies. but the people who do the actual work. are reporting significantly lower numbers. Governments and companies may be demanding a world with zero vulnerabilities But that land, feels out of reach.
[11:18] given our tools and techniques. The sobering bottom line is that we are no more secure today than we were back in 2021. The silver lining, however, is that we have woken up. And we are more aware of these challenges than ever before. For us at Slim, we believe you should know what's inside your container. and you should ship only what's needed to production. We're seeking to solve this challenge for all of the world's containers through automation and intelligent optimization.
[11:54] And look. I do not think complexity is the enemy here. Ignorances. There are times where we need to take a step back and enjoy the poetic beauty of the complex systems around us. There are other times where we need to step in fearlessly and deconstruct and redesign. I know.
[12:26] that we have to resolve the decision intelligence. the willpower to solve these issues. And most of the decision makers are in front of me. In this room. My friends say I'm an incurable optimist. But that's not why I think the future is bright. I know there are relentless Brilliant! competent teams around the world who are losing their sleep thinking about these problems.
[12:58] So many teams. that I couldn't even fit. the names that we leveraged for this project alone in my thank you card here. And as humans, when we gather, like we did at this fabulous location When we put our hearts and minds together, We can make even ridiculously tough beats Lupeze. And that's why I know when I'm on this stage next time, I will be bringing.
[13:30] Better news. Thank you. ♪ ♪
在Vidleaf工作台中打开
搜索转录文本,选择行,复制带时间戳的引文,翻译。
署名
"主题演讲:我们在剖析全球最受欢迎的容器时学到了什么 - Ayse Kaya",由 CNCF [云原生计算基金会] (https://www.youtube.com/@cncf) 制作,依据 CC BY 3.0 (https://creativecommons.org/licenses/by/3.0/) 许可。源视频:https://www.youtube.com/watch?v=f2-nk5qIur8。本页面是该视频的文本转录,添加了段落分隔和时间戳;创作者与 Vidleaf 无关联,也不为其背书。Vidleaf 添加了 AI 标点和 AI 生成的摘要;原始措辞仍可在本页面上查看。
您是创作者或权利持有人吗?请求更正或移除:copyright@vidleaf.app(参见关于这些页面)。
最后更新于