Hacker news

  • Top
  • New
  • Past
  • Ask
  • Show
  • Jobs

How to Break into Security, Ptacek Edition (2012) (https://krebsonsecurity.com)

128 points by ronfriedhaber 4 days ago | 32 comments | View on ycombinator

liquid_oil 4 days ago |

Reading a post from 2012 like this it's shocking how quickly the ladder has been pulled up in all computer careers, but I suppose it's not as bad as the ladder pull from the dotcom bubble to 2012 where you could previously land a career "knowing HTML".

sorentwo 4 days ago |

Marking this as 2012 would help. It’s a fun read in a nostalgic, pre-ai kind of way, where emphasizing that developers should use bcrypt wasn’t standard practice.

bryant 4 days ago |

tptacek may disagree, but this needs an update considering how much has changed both with AI and the economies of scale AI brings to the discipline.

I'm anticipating "the fundamentals are crucial," and as much as I want to agree, I think practical realities show us otherwise. We're at a point where all someone needs to know is how to AI-program-manage and how to glean enough about the discipline (or any discipline at this point) from AI to be dangerous enough in the field. To be a builder, breaker, or defender wearing any color hat of their choosing... that's it, really.

smashandgrab 3 days ago |

I hire for incident response roles. Honestly, despite how much has changed this is still the way to do it for us. You need to be versed in many things and more importantly, interested enough to not feel entirely stressed out when things hit the fan.

As for AI it’s been great for things like log review but those were already pretty low skill requirements. We haven’t had much luck past that.

michael-bey 3 days ago |

It's still pretty good advice. I would replace "learning how to program" with "learning how to critically evaluate LLM output".

Like almost everything these days, it's K-shaped. High agency people who use LLMs as a tutor will be able to more quickly break into security. The script kiddies will still script kiddy, except instead of running a random Python script it's doing it uncritically from Claude / Cursor / Codex.

As someone who had to break into security myself around this time, I would have loved an LLM instead of trying to cobble together understanding from forum posts and PDFs.

cyberax 3 days ago |

What is even security, these days?

Algorithms are a solved problem. Post-quantum encryption was the last bit of excitement for a while. I guess ZKPs and FHE are still available, but they are really niche.

It's everything else that fills me with dread: we're using oodles of software written in C/C++, so we're one missing array boundary check away from the global economy collapsing.

Then there's phishing and social engineering. If one of our developers gets compromised, it's trivial to push malicious code into our repo. And from there it's an easy escalation into our customers' billing, private information, etc. We store the bare minimum of data possible, but it's still a lot.

j-scott 4 days ago |

(2012)?

tptacek 4 days ago |

At this point, who knows?

Arch-TK 3 days ago |

Security pays well, but software engineering still pays better. Even significantly with the right luck and skills.

dzonga 4 days ago |

it would be interesting to know what tptacek's thoughts are in the age of 'agents' violating vulnerabilities willy-nilly ? if his thoughts of getting into application security still apply.

what the incentives are since hacking seems to be no longer a crime given how many hacking incidents have gone unpunished.

chias 3 days ago |

Does anyone want my spot?

david_shaw 3 days ago |

I'm often asked how to turn a security-related interest or hobby into a job. For the longest time, I'd say pretty much the same thing written here: get a book, do some research, build some projects, talk to people in industry. I've been going to DEF CON for about 25 years now, and I've had this conversation countless times.

But in 2026, tptacek says it best in the comments here: At this point, who knows?

Are we still telling kids to learn programming languages to become software developers? Perhaps system design is a more reasonable skill set.

That said, because this is the time to say it: security is still important, and it's not -- and shouldn't be -- fully automated.

The toolchain is changing just like it always has. I remember when interpreted languages replaced compiled ones, and people said that interpreters solved security. Then people went from writing their own codebases to implementing frameworks, and people said it solved security.

Barracuda Networks posters at every airport stated "Security Solved."

As long as people take risks, security is important. Now we just need to focus on designing resilient systems instead of finding bugs and vulns on the code-level. LLMs can do that part for us.

silentsobs 3 days ago |

[flagged]