Hacker News Reader

A few years back, on day 1 of my new job I was given root access to one of the development boxes.

So I ask: "Okay, how do I log in?"

The IT guy: "What do you mean, you just log in using your personal domain account and then sudo su -. You know what sudo is?" (followed by loud sigh)

Me: "You mean like production domain, same that we use for our desktop?"

IT guy: "Of course! What do you mean, what other domain would you like?"

Me: "Can I at least change my password to something else just for the dev environment? Can I log in with SSH key?"

IT guy: "No, no, no. Per our SECURITY policy, SSH keys are disabled and you have to use our domain login and password". (another sigh... of course)

Me: "Are you aware that when somebody has root access to the box they can do whatever they want including intercepting passwords of all users that log in to that box? In this case, every single developer that ever needs access to dev environment?"

IT guy: "That's not true. SSH is encrypted protocol and it is not possible to access passwords".

Me: after many tries to explain this to various people from IT, I gave up and set out to intercept all passwords of all IT employees. After I had passwords of almost everybody, I put them all in an excel and sent to IT for "verification".

There were a lot of angry people that day wanting me fired... fortunately they came to their senses.

Unfortunately, my development box access privileges were revoked.

I once decided to show the vulnerability of SMTP protocol by sending an email as a higher-up. (Too young, too naive, don't ask why I did that.) Created a massive firestorm. I did successfully convert them to use SPF and DKIM and showed everyone the need to never trust an email. Some even adopted PGP signatures after that.

I like to imagine you sent an email as a CEO saying “please read this ASAP” and then it links to a Rick Roll

> There you go, giving a fuck when it wasn't your turn to give a fuck.

>

> - Bunk

"The standard you walk past is the standard you accept."

I think even if you were fired, you'd have dodged not a bullet but a cannonball there.

Sometimes being the nice guy full of good faith doesn't pay. Literally

Nah, I worked there for a long time and I was sad when I needed to leave. I had best boss and best project ever and I learned and grew a lot.

That's good to hear! Although I would have been frustrated and made anxious by their security practice. Did they change their security practices after that 'event'? And did you manage to regain access to your development box?

>sudo su -

It's high time for this thing to die. sudo supports this natively since forever:

  $ sudo -i

Why

Curious, whats the difference and why is sudo -i better? I was unaware of sudo -i (I might as well be a noob though), so I had a quick search to find out more and found this [0]

One question, if I may, could there be any situation (that you can think of, off the top of your head) where using sudo -i would be worse then "sudo su -"

[0] - https://www.maketecheasier.com/differences-between-su-sudo-s...

'sudo su -' uses root to run 'su -' to open a root login shell.

'sudo -i' opens a root login shell.

su is old fashioned but it works fine. I don't even have sudo installed on my PC.

Huh. TIL.

I usually just do `sudo zsh` :-)

Wait, seriously? How would you do that? And why doesn't it work for ssh keys?

I was under the impression password auth worked very similarly to key auth, as in, the actual bytes of the password or private key are never sent over the wire.

No, password auth transfers the password and (usually) hands it to PAM on the server to verify - the server sees your password. Key auth happens in SSH and the private key never leaves your machine.

(If I remember correctly just putting SSHd in the highest debug level, which logs every byte received, did work to intercept passwords)

On the plus side, it also gave you plausible deniability it really was you if you wanted to do something illegal or unethical.

chaotic hysterical

Like jeez, GP. Should’ve sold those aerospace secrets to China while you had the chance.

Pro tip: When you're seeing something really unethical that could eventually rebound on you - write it down. Write down contemporaneously what happened and sign and date it. Much more reliable evidence.

Send it in an email too. "Hi IT guy and Boss, I'm concerned about blah, but understand people need access to everyone else's computer and account. Yada yada." Don't be difficult, but get an email out there with a time stamp on it and forward all that to your personal account if you can.

Any UPS store or FedEx should be able to notarize it, in the unlikely event there isn't a registered notary among your coworkers.

Bonus points for writing it digitally and using your favourite proper or janky method of timestamping - dates on paper are meaningless, even with a signature (which is also meaningless itself).

Emailing it to someone seems to be the go-to recommendation online, but you might not want to expose the information at the time. Some alternatives, in descending order of jank: a Google Doc, unlisted Pastebin, either of those but only post the hash and keep the file offline, a proper TSP timestamp - the latter is actually near-trivial these days with something like freeTSA.org.

The funnier thing about that story is that the passphrase they gave you is not necessarily secure in the first place if we're talking about a scenario where a hash is found. While length is a factor when brute forcing randomness, any words or phrases that commonly appear together in written text are likely to come up in various types of dictionaries that can be used in more sophisticated brute forcing algorithms.

"sliding down the" and "tall building" are two parts of that password that contain words very commonly found together.

It's not the worst password but it's better to introduce randomness.

One strategy to create strong passwords that you can remember is to pick 3 or 4 truly random words from the dictionary that have no logical connection to each other, and to "glue" them together with some numbers that you can remember. I've heard that phone numbers are 7 digits because people can remember up to 7 pieces of information. If that's true then most people should be able to remember a password in the form of: <word1><number1><word2><number2><word3><number3><word4>

It does start to break down when you have to remember multiple passwords. So using that format as the master password to a password manager and then letting your password manager generate truly random passwords for everything else is the way to go.

The nice thing with the lists of multiple randomized words is that there are so many more combinations possible. The numbers, which will be individual digits for most people, add relatively little here.

Using EFF's dice-ware page (https://www.eff.org/dice) as an example, the long word-list has 7776 entries, with a nicely random way to select them. Four of those words is already a sixteen-digit number in the combinations available. Even their short list of 1296 words would provide pretty reasonable odds. The key is to use something like the dice that takes away our own biases when selecting words.

Really the key is using words that aren't in a dictionary, or at least not the same dictionary. It's pretty trivial for a brute force attack to just take a list of English words and try them in a large number of combinations. But if you have some non-english words, as far as that attacker is concerned its random gibberish. Of course the attacker might use a few languages so just mixing say english and spanish won't do that much better, but if you got some navajo or some quenya you're likely immune to any attack that isn't specifically targeted at you. And it doesn't even have to be a foreign language, even many proper nouns won't show up in a dictionary. For a password like "PoggleExclaimsGeonosianBarahunde" or "Seanbeanfearsburzum-ishikrimpatul" it would take trillions of years to guess a password without very specific pop culture knowledge.

Here is a small back of the envelope calculation. Websters dictionary includes 470.000 words [0]. 52 lower and uppercase letters + 10 digits + ~20 special chars = 82 possibilities. There are are 82^20 = 1.89 * 10^38 possible combinations for passwords consisting of 20 random characters. Picking 7 random words from Websters dictionary has 470.000^7 = 5.07 * 10^39 possible combinations. I'd argue that remembering 7 words is easier for most people, than remembering 20 random characters.

[0]: https://www.merriam-webster.com/help/faq-how-many-english-wo...

Yeah, for a given level of entropy words are a better choice, but the fact is that dictionary attacks greatly reduce the efficacy of what could be an incredibly strong password if they can safely assume it is formed exclusively from concoctenated standard words. However if you break the validity of that assumption by using a word that's not on a list, then the dictionary attack must be combined with a brute force attack with the same string length. If we assume an average word length of 5 letters, that 7 word password jumps to 1.22x10^79 possible combinations if you randomly substitute one letter.

My work is also big on pass phrases but I kinda hate them. It's a lot of extra work to type it every time I unlock my PC. And more characters means there's more chance to make a typo meaning I have to do the whole thing again.

And the weird way my brain works I have no issue remembering "G6bH,vIz#amV" so I still do it like that :)

Also, if an attacker has a hash of my password the damage is already done anyway. They can do a pass-the-hash attack, they don't even need to brute force it for the plaintext. So there's no real benefit there. What matters is online guessing and that is severely limited in the amount of attempts.

I can't wait for passwordless though. I use only smart cards/yubikeys at home and love them.

> Also, if an attacker has a hash of my password the damage is already done anyway. They can do a pass-the-hash attack, they don't even need to brute force it for the plaintext. So there's no real benefit there.

A diceware password of 6 words has an incredible amount of entropy. How would you ever brute force its hash?

> What matters is online guessing and that is severely limited in the amount of attempts.

Belt and suspenders.

In pass-the-hash, the password doesn't matter. If you don't know the hash, it's equivalent to online brute-forcing a password of a fixed length, but since that length is usually more than the password itself, it's usually better breaking that. An NTML hash for example has >3× the entropy of a 6-word diceware password, but in an insane situation like 10-word diceware vs md5, brute-forcing the hash would be faster.

What's stopping you from using a random password like 'G6bH,vIz#amV'?

Well they wanted to set the minimum length at 16 characters, for one :)

But in the end they settled on 10 so it's ok. They even made it mandatory to have special chars and numbers if you have less than 16. So in the end it worked out fine for me but it did take some convincing (I was involved with the team that was making the choices).

Tbh even if you do have to brute-force, if you know the company is using passphrases, it tends to be similar in terms of difficulty as a complex shorter password to brute force it. Because you can do a combined dictionary attack. And it's more susceptible to targeted attacks e.g. gathering a person's interests on facebook.

> They can do a pass-the-hash attack, they don't even need to brute force it for the plaintext.

Ok, but that's a very Windows-specific issue. I don't know of any other widely-deployed system vulnerable to pass-the-hash.

That's true but we're a Windows shop, sadly :)

I'm a bit appalled at the security of AD with PTH, Kerberoast etc. In some cases you can even continue to use a nabbed ticket after the compromised account has been locked! That should never be possible IMO.

I'd love to move on from AD personally. But you know... Legacy galore.

Wow I know that password expiration is on its way out [1] but this is crazy.

[1] - https://www.sans.org/blog/time-for-password-expiration-to-di...

Makes you wonder if you could use a tool like GPT-3 to generate grammatically correct passphrases from randomly generated sets of words.

E.g. "correct battery horse staple" might become "Correct use of a battery supports the a horse in its staple diet".

Thereby keeping the same bits of randomness while making a memorable phrase.

It also gave you plausible deniability which would have stood up in any court.

Well, was the Excel spreadsheet accessible to everyone? If not, this model could work in some strange way - if someone needs access to your computer, they are provided the password, then it is changed and you and password custodian now have a password not known by everyone else.

One side effect of this is that if you know someone else has the password, you're probably very unlikely to do any personal business on that machine.

There has got to be a better way, namely, not an excel spreadsheet (maybe vault or LastPass) and not the individual employee's account (e.g. have an IT super user account).

I think OS miss the ability for an admin to log into a user context. I am very hardline defender of privacy and in principle I support privacy advocates already preparing my pyre for even suggesting that of course.

But if we are honest an admin can already access relevant data anyway and you do indeed need to use a user context to access certain settings of some applications. This is especially true for less digitally affine users.

There are workarounds (Windows->Run as...), but that is sometimes insufficient.

About an admin using your user account to start the nukes? That is mostly possible anyway in the usual corporate Windows configuration. He could just change your AD password and log in as you. I would recommend that approach anyway and then force the user to set a new password and informing him why you sabotaged his workstation would also be nice. Far more secure than having the holy grail Excel file.

The first place I worked was like this, with the exception of the long complicated passwords. We told people to call the helpdesk to reset their password and would save their password to an Excel sheet. This was improved slightly by rolling out a self-service password change tool that logged the passwords to a database.

One day someone that had legitimate access to the password list was fired/quit and the VP decided we needed to have everyone (~800 users, 700 remote) call the helpdesk to reset their password. All the helpdesk calls waiting for a password reset clogged the voice T1 (23 lines) for the office, preventing most calls in or out. Oops.

Thankfully when the company came in-scope for sarbanes-oxley the external consultant we hired to help with the audit said there was no way we could continue logging user passwords and stay compliant.

This level of negligence should be criminal.

The software industry is full of should-be-criminal forms of negligence.

Things are already horrendously bad. Basically every American's identity could stolen at this point. If any nation state or other actor decided to operationalize any of the big leaks -- eg OPM or EquiFax -- the ramifications would be catastrophic. Imagine millions of people losing their retirement accounts and all their savings. Even if you could correct everything -- and that's a big if -- the process might take years and the intervening panic would be deafening. The amount of anger might even elicit a hot response.

To say nothing of more serious vulnerabilities. We really dodged a bullet on the pipeline ransomware.

I'm morbidly curious how bad of a "Cyber 9/11" we'll need before software starts being taken seriously as an engineering field in which practitioners have professional responsibility.

So you’re saying that is a gigantic target for China and Russia to go after lol. It would mean some change for the which might not be bad but that’s kinda like arguing for terrorism, which would be illegal and actually have enforcement behind it.

No, it's far too large and undivisive a target for China or Russia. They're both too smart for that. They fuck shit up in ways that are partisan and divisive.

This is more of an NK/non-state-actor "burn it all down" move.

I would assume that a few nation state actors have already hoovered up all that leaked information and have it implemented in a system that is ready to steal identities, drain accounts, and otherwise wreak havoc on a large scale. They are just waiting for the higher ups to pull the trigger if and when they decide to deploy it.

Yes. I'd be astounded if multiple such attacks aren't ready to deploy. At least 6? Maybe as many as 12. What it's waiting for is a desperate enough actor or a weak enough moment. Russia and China learned from Japan and ISIS et al learned from bin Laden -- divide, don't unite.

But eventually there will be a sufficiently naive actor and/or a sufficiently weak moment.

The tragedy, of course, is that all we have to do is address the totally and completely obvious problem. It wouldn't even be that hard. But we won't. Last 4 of SSN is still enough to transfer a SIM even with explicit direction otherwise, and transferring a SIM is still enough to drain a bank account even with explicit direction otherwise.

Eh I think you went into hyperbolic assertations here. Might want to walk it back. It's bad but it's not as bad as you say.

Suppose NK or some non-state actor has access to EquiFax or OPM data. Or just a conglomeration of various other data leaks, some of which are probably still unknown or at least unannounced. Suppose that actor launched a plan with trivial lead-time and resources -- say, 1-3 years with 10-50 trusted primaries and however many in-the-dark "contractors".

Do you really think the result wouldn't be turmoil for at least weeks, if not months, and take a year+ to unwind? Serious question. If so: please explain.

if that were to happen it would be a "to big to fail" event. the gov would bail everyone out and mandate a reset to what ever it was believed to be before. bank accounts, retirement accounts etc would be reimbursed up to the FDIC limit. credit reports would be rolled back to the last known good value. it would probably fuck things up short term but long term it would lead to better security of consumer data

Completed agreed. But the short term fall-out would be incredible, and I think the backlash against tech unprofessionalism would be merciless.

I really like that term “Cyber 9/11”. Is that something you made up or is that a term people use describing a bad attack?

Thanks, but I can't take credit :) I think I first came across this term in a foreign affairs article, and from then on used it often in presentations to various brain-dead military officers who all seemed to have degrees in Biblical Studies from colleges whose boards are full of Domionionism types.

In any case, the term has been around for a while. Unfortunately, our officer corps is populated by weak-minded fools who have more allegiance for their quasi-Baptist cults and podcast hosts than their country. They all seemed to have a multi-year education in how to use Hebrew language factoids for isogesis, but had no god-damned clue what a "heap" was, and were effectively mid-level managers of "Cyber Operations"...

Our current state of affairs in the civilian sector isn't too surprising and I have infinitely more confidence in random banks and credit "borough" companies than I do in our military.

All of that to say: a cyber 9/11 attacking civilian infrastructure is a best-case scenario because that's where all the good people are. An actual 9/11 will probably attack the defense sector where all the incompetents work and will be way worse than actual 9/11. You have officers with theology degrees from shit-tier southern bible colleges to thank for it. After we're done bombing whatever rural town the hacker happened to live in, our next two steps should be professionalizing software engineering and writing history books about how christian fundamentalists destroyed the integrity of the US officer corps.

I didn’t realize there was so much religion in the military. Gun in one hand bible in the other is the dumbest thing I can think of. Do they actually think if heaven and god was real they would be invited? Ha! I would love to hear how a military man justifies going to war and then preaching the bible surely the two go against each other.

Prepare for failure. A good rule, but painful is: The more income tied to an account the greater the difficulty to move the income. I'm too tired to list best practices but for example: Set up canaries, daily emails from your account just for the peace of mind that your email is the primary communication for the account. Biggest assets should take time and multiple steps to transfer or cashout. Know your account managers and be able to contact them directly.

Hi, I hope this doesn't come across as me not respecting your tiredness but could you link or anything to best practices if you don't have the energy to write it out yourself?

Well, I will try. First, I was commenting on my interpretation of the parent comment. Identify theft, and protecting savings and retirement accounts.

Be inquisitive and aware. I think you have this covered by reading hacker news and having an interest in the subject. I've enjoyed reading Slashdot (while it was good) before switching to hacker news, but it's also been a vital ongoing education for me. Comments often having more value than the original article. Being knowledgeable of security risks and common exploits helps prevent falling victim to them.

https://hn.algolia.com/?q=identity+theft

https://www.newyorksecuritieslawyersblog.com/my-money-was-st... Good read on how someone lost their account.

Steps for securing accounts. Confirm that you are notified of email address changes. Confirm that you are notified of any transactions on the account. Setting up a canary if possible. I set up an email alert on a common event. So, I basically I get an email from the company daily, and this confirms that my email address has not been changed. If you are certain you will be contacted if your email address changes then this is not necessary as the email change notification acts as the warning. Have email and phone of account representative that you can contact if there is a problem with the account.

That should be all that is necessary. Now, the day comes, someone has changed your email address. Maybe they even did some transactions. Stay calm, stay professional. Contact your account representative and notify them of the problem. Be able to identify yourself, call from a phone number associated with the account (or previously associated). Be able to answer security questions. Account representative should be able to freeze the account and resolve any issues. If you're satisfied with the phone call, great. If you're in anyway nervous about the resolution, then create a paper trail, send a letter that documents the issue and your attempts to resolve it.

A quick disclaimer, I'm not an expert. Adjust anything to fit your own needs.

Would you mind clarifying what a canary is? I understand it is referencing the idea of a canary in a coal mine but what does that look like in this context? Other than having a separate account to which you transfer most of your funds so you can't get robbed at gunpoint and be forced to transfer someone your money I'm drawing a blank here - sorry.

A canary is a safeguard against dangers. In my example, my daily emails from my investment account is my canary. When I stop getting emails, I know there's a problem.

I cannot remember the details, but one of my favorite canarys was a website that had a paragraph that basically stated we have not been compromised in the past 24 hours. It had a timer that had to be rest daily or the paragraph whould disappear from the website.

The slow guy that still kept paper records would win this.

I don’t see why. I think it’s generally silly to think of your work computer/account as a place for anything private. Even if they won’t normally look at it, it’s all fair game if e.g. someone subpoenas them. My employer had a team which did a similar thing to the GP (had a spreadsheet with everyone’s password) in case someone was out and had some crucial file on their computer. And while they have since stopped doing that, it is because they improved processes such that it wasn’t needed and not because it became acceptable for the team to fail to carry out a large portion of their business activities or obligations because someone important was sick.

The problem is that the spreadsheet meant that exploiting one user meant exploiting the whole team but you can have this kind of privilege escalation in plenty of other ways. An easy way is to give people lots of permissions so that they can get their work done, or to just be bad at revoking permissions once they are no longer needed. Plenty of companies deliberately give people wide read permissions as a part of a culture of openness.

It's not about it being private - clearly the business can have root/admin/domainadmin/whatever access.

However your username and password identifies you. If user "johnsmith" does something, then that's because johnsmith has logged in. Now IT may have changed the password to allow them to log in as johnsmith (either following some odd policy, or a rogue IT worker), but that would be in the audit log.

If a company needs more than one person accessing an account for some reason, they should create a generic account (e.g. "z_reception" for a generic reception machine).

Yeah, it's the similar in the company i work for. Except that we have a mixed bag of passwords. There are some old ones that are like "mouse" and then there are others like "Gt4x;JlK". We have that list for "maintenance reasons". Doing things, that could easily be done by some GPOs and other cool tricks.