[devnexus 2022] hacking the OSS supply changes

Speaker: Stephen Chin

@steveonjava

Link to table of contents

———————

Theme is security with sci fi references

Examples

  • Equifax data breah – from not patching Struts for at least two months
  • Solarwinds – hacked TeamCity instance injected
  • log4shell – zero day in log4j core. Affected almost all systems. Could send class file and having it excecute on the serer
  • spring4shell

Binary repos

  • Which do you trust?
  • npm, pypi, rubygems, maven central
  • Like picking up thumb drive off sidewalk and plugging into your production server

Dependency confusion attack

  • Sci fi – Matrix – agents disguised theselves as other people
  • package mining
  • npm has no security on namespaces
  • Can use same name as a company internal package and give it higher version number
  • If grabing latest version, pull mallicious package
  • When pull from npm, announcing what package you have
  • Artifactory resolves against internal repo first. Protects even if using virtual repo which mixes public and private content

Supply Chain Attacks

  • Sci fi: millinium falcon
  • Assume depedencies built on a clean system
  • Anyone can upoad to pipi
  • About 400 zero day volunerabiities in open source/cloed source/OS, embedded systems, etc
  • Sveder uploaded library to go to his website
  • JFrog scans looking for suspicious Python code behavior
  • noblesse – “optimizes your PC for python” – steals credit card/passwords and sends via dicord
  • pythatoras – supposed to help with calculations but does remote code executio

Namespaces

  • Sci fi: War games
  • Moscow – Russia and Idaho
  • St Petersburg – Russian and Florida
  • azure-core-tracing is proper name. Created core-tracing.
  • NPM took down once repored. At least 218 packages affected.
  • Stole personal data
  • Think bug bounty of test because minimal and not steaing credit cards

Pyrisa

  • Scitfi: Avengers
  • Need automated (IronMan), trustworthy (Black Widow) and dependable (Captain America)
  • trusted binary network – secure by defaut, reliable inimal outages), open
  • peer to peer
  • multi-node verification
  • reproducabe build trust model

Websites

  • research.jfrog.com

My take

I hadn’t heard of all those attacks so learned about the Python ones. The sci fi element was a nice touch. As was the community picture with a ton of people on stage.

[2020 devnexus] Developers Need to Take Notice – Malicious Attacks On Open Source Are Getting Worse

Speaker: Derek Weeks @weekstweets

Deck: https://speakerdeck.com/derekeweeks/devnexus-presentation-derek-weeks

For more, see table of contents


Fun

  • XKCD – reinvent the wheel – “We don’t want to reinvent the wheel, so every day we google image search “wheel” and whatever object comes up, that’s what we attach to our vehicles. Sure external dependencies carry risks, but so far, they’ve all been pretty good wheels”

Notes

  • If just moving fast, have a problem
  • DevOps teams use more open source
  • 70% deploy at least once a week
  • Challenge: be faster than evil
  • In past 5 years, breaches increased 70%
  • Can’t predict when vulnerability will come up. Have to use without knowing what will happen.

Attacks

  • Equifax is old news by now. Had opportunity to patch
  • Adversaries can also contribute to open source. ex: npm event-stream attack on CoPay
  • 2019 – Gems bootstrap-saas – added backdoor
  • Typo squatting
  • Backdoors
  • Happening to Docker, Python, Ruby, NPM, etc

Supply Chain

  • 2019 Software Supply Chain Report
    • suppliers (open source)
    • warehouses, (component repos)
    • manufacturers (softare dev teams)
    • finished goods (software applications)
  • Maven Central had over 200B downloads in 2019 alone. Almost 10% had known vulnerabilities they day they were downloaded.
  • JavaScript hit 10B package downloads in 2019 alone. Just over 51% had known vulnerabilities they day they were downloaded.
  • 85% of app is sourced from external suppliers

Enterprise vs Open Source

  • Multiple deploys per day vs versioned releases
  • Consistent Dev team vs fluid group of developers
  • Predictable/well resources vs variable resources
  • Deploymen tvs release frequency
  • Organizational performance vs popularity
  • Mean time to restore vs time to remediate vulnerabilities

Is it true?

  • TRUE: Projects that release frequently have better outcomes – more popular, attack more developers and higher level of support from foundations. Also, avoids problem of having to wait for all the transitive dependencies to be in a version we are using.
  • TRUE: Projects that update dependencies more frequently are general more secure
  • FALSE: Projects with fewer dependencies will stay more up to date. Interestingly, smaller teams tend to have less dependencies. Not clear if correlation or causationl
  • FALSE: More popular projects will be better about staying up to date

Behavioral clusters

  • Small teams; excellent time to update
  • Large teams; excellent time to update, often foundation supported, popular
  • Laggards – release slowly, more likely to be commercially supported
  • Features first – release frequently, but poor time to update
  • Caution – good time to update, but seldom completely up to date

Developers

  • 38% schedule dependency updates
  • 46% strive to use latest version
  • 50% have process to add new dependency
  • 30% have process to proactively remove problematic or unused dependency
  • 37% have automated tool to track, managed and/or ensure policy compliance of dependencies

If only do one thing

  • If stay on latest version, by default more secure and less security issues.

My take

Good talk. Especially once the projector issues were fixed. I lik the graphs and data behind the main points. I’ve seen similar presentations, but the newer parts/stats were still good to hear.