In the TLS handshake: "The server generates its own hash, and then decrypts the client-sent hash to verify that it matches"

The server decrypts a hash? But thats not how hashes work.

The client encrypts its hash before sending it to the server. Thus the server must decrypt it to compare it to the one it generated.

Hashes are one-way, so cannot be decrypted. The server can _compare_ the results of a hash (by doing the hash itself, and comparing the results), though.

You and turtles are suffering from the cryptographic equivalent of a hypercorrection, in the same way that well-intentioned people insist on the propriety of the grammatically impossible phrase "between you and I" (which should be "between you and me," because prepositions take objects, not subjects.) The two of you have had the irreversibility of one-way hashes drilled into your heads, just as many of us were taught when young not to say "me and Susie were playing on the swingset." And you have an allergic reaction to anyone using "decrypt" and "hash" in the same sentence, which can lead to that allergy triggering a false positive. In this case that's what's happening.

Cryptographic hashes are irreversible. That's the point of such a device. But there is nothing stopping someone from taking the result of a cryptographic hash and then encrypting it, and then that someone or someone else decrypting that ciphertext to recover the hash result. E(H(S), k) leads to an encrypted hash, and D(E(H(S), k), k) recovers the hash. It's computationally infeasible to retrieve S. But nobody wanted to do that; they just wanted to know H(S).

You are correct that the server compares the result of the hash (which in context can also be called a "hash," such as "I used SHA-256 on my term paper, and then I spray-painted the hash on the face of the town clock tower, thus proving the existence of my term paper before the class deadline"). Nobody's arguing that. But how did it obtain the thing it's comparing its own result to, without M also obtaining that thing?

(I'm actually not sure whether TLS sends the actual hash or bases subsequent computations on the assumption that both sides can independently derive it. But if it does the former, it's totally fine to say "it decrypts the hash," which is the objection of the parent of this thread.)

Thanks, TIL :)

I can create a SHA-256 hash, then encrypt it with AES. You would then have to decrypt it to read the hash inside is what the parent is saying.

Most of that description is outdated and/or wrong. Probably this HN article should say (2015).

But yes, of course you can decrypt an encrypted hash, this way you get back the plain hash.

The client calculates a hash, it _encrypts_ that hash, and sends it to the server, the server _decrypts_ it, and then can verify that it has the same calculation.

The reason this is done is that it can detect a situation in which the client and server were persuaded to arrive at the same results by different means, whereupon they should abort the connection. The mechanism in TLS 1.2 and earlier was not very good, a better one is included in TLS 1.3 but alas last I looked it is disabled in popular browsers because it's incompatible with yet more middlebox crapware from "security" companies.

It should read "then compares the generated hash to the hash that was received from the client to verify that it matches"

Or something along those lines.

I assume the reason for doing this is to confirm the symmetric key now in use is known between both parties?

I wrote it above, but more relevant here maybe: No. There's no need to confirm that, if the keys don't match everything will fail anyway and the connection aborts because everything either party sends appears to be gibberish.

The description linked over-simplifies, the hash they're calculating is a summary of the handshake process by which keys are agreed, we want to prove that both saw the _same_ process happen to reach this state.

Suppose I am willing to use archaic method A because I'm a simpleton, although I do know methods C and E which are safer. The wise people running www.google.com only allow method A if you don't know methods B, C, D or E.

Now, I try to connect to www.google.com and unknown to me a Bad Guy is in the middle. I say "Hello, I know methods A, C and E", but the bad guy changes that message to say "Hello, I know method A only". Google replies "OK I guess we can do method A then" and we use method A. The Bad Guy knows how to break method A and now my security is ruined!

But with this Finished message in TLS, www.google.com and I will calculate different hashes, since I know I said "I know methods A, C and E" but www.google.com got a message from me saying "I know method A only" and those don't hash the same.

This proves somebody is tampering with our connection, we must abort.

ah, cheers.

“30 seconds”. You should have told him 6 hours. If he needs it done quicker hire someone else. Enjoy your day.

He wasn't able to successfully pull this stunt ever again. It actually completely changed our work dynamic and almost completely eliminated job stress. It was also the catalyst to deciding that I wanted to move on and I gave a 4 month notice shortly thereafter. Ultimate jerk move, he actually hid a bonus check intended for me in his desk for weeks because he didn't realize I was being so nice with the amount of notice. He thought he could prevent me from getting it entirely. That did not go over well with the president of the company.

So, theft and fraud eh? Sounds like a lot more than a "jerk".

Hopefully, the president canned him and referee his conduct to the police? Seriously, the less poisonous people like that, the better.

Yeah. I'm also like wtf. The website looks about 10 years old.

I assume the owners have good communication, and persuasion skills. Which is admirable, since I lack them.

Hi, this is a bit of a random one, but I have product that relies on GPS accuracy, and would love to hear if you guys have any ideas on how I can improve this, as well as my current implementation. Mind if I drop you an email? Cheers.

Had been thinking of moving to CA, but might as well stay in Sydney by the sounds of it.

Perhaps, but you might learn more in CA; bigger market, moves faster, more experienced co-workers, stronger culture of entrepreneurship, generally bigger thinkers.

Worst case scenario is that after earning USD for 2 years on an E3 visa, you return home more experienced than your peers with "Silicon Valley Startup Experience" in your resume, if that is something that appeals to you. Stronger startup ecosystem on the small scale, bigger problem sets on the large scale.

Don't get me wrong, the startup and tech scene in Sydney has grown into a solid network over the last 5 years, but you have to weigh that against the certain knowledge the Bay Area accounts for 40% of U.S. venture funding and even that pales in comparison to the tech giants who are either based or have sizable offices here.

In terms of lifestyle, Sydney will always win. I suspect this is why there is such a strong loop back home after Aussies spending some time here.

Is there a possibly for the Senior Security Engineer role to be remote?

This is not the type of Uber driver I want. I don't want someone complaining about their life choices when I'm paying for an agreed service from A to B.

This is a big problem with taxi drivers is australia. At 2am you'll be lucky to get a taxi home, because they change over at 3, so they don't like to go 10 minutes in the opposite direction.

I'm glad there is the Uber alternative!

It turned into a safety issue too with stories about women being stranded because the fare was 'too low' or 'in the wrong direction' for the driver

A junior pentester as remote, thats risky.

Can someone please ELI5 why this is good, and what they can be used for? I'm assuming machine learning...

I haven't anything with fastText, but I have with word2vec. It embeds each word in a 300 dimensional vector, such that similar words have a large cosine similarity. (If you normalize each vector to have a unit norm, then cosine similarity is just a dot product.) So in short, it gives you a measure of how similar each word is to other words.

This has many uses in machine learning. You can extend it to documents and find similar documents, find misspellings, use them as features in a ML model, etc.

There haven't been good vectors in that many languages (that I know of), so that's a plus for these fastText vectors.

ah. Thanks!