The Real Problem with the Twitter HackCrash

In case you missed it, the @AP twitter account was hacked , which resulted in a tweet that sent markets spiraling down only to recover a few minutes later once it was revealed that the tweet was a fake. Here is more info.

Why did the market head lower so quickly ?  This is from Paste Magazine

“The events last Tuesday were likely caused by the news-reacting algorithms that are designed to electronically read and interpret machine-readable news,” said in an emailed response by Irene Aldridge, a hedge fund consultant on algorithms and author of High-Frequency Trading: A Practical Guide to Algorithmic Strategies and Trading Systems. “Most of the web content is machine-readable, so lots of algos are built on reading and reacting to news such as the Twitter hack… Clearly in the Twitter case, trading machines did not ascertain credibility of the tweet,”

So how does the future of HFT look in light of Tuesday’s AP hack tweet incident?

Aldridge said in terms of the technology, “going forward, many algorithm designers will take into account the Tuesday situation, and will build ever more sophisticated approaches.”

There is no question that there is an ongoing arms race between algorithm designers, as well as a technology race to improve Language Processing and Translation abilities (disclosure, im an investor in Linguasys ).  Billions of dollars are being invested to make trading without humans faster, cheaper, smarter. The problem is that no matter how smart you make machines, they will never be smart enough in our lifetime to detect all levels of deceit and fraud. Particularly online.

No one has any idea, other than the traders using the algorithms, which twitter accounts the algorithms follow and read. The same with websites. Which do they track and read ?  If its worth it for someone to hack the @AP twitter feed in order to attempt to destabilize the markets why wouldn’t the same people or others with similar goals set up thousands of twitter accounts that for some extended period are a solid source of news and information, attempting to build  a following that include algorithms reading their feed , knowing that at somepoint they will tweet market moving fake information ? (Note, you dont have to actually follow an account to get access to their feeds, there are services that provide the twitter hose to financial companies)

It costs absolutely nothing to try to make this happen.  If you follow the writings of Nanex, you know that mini flash crashes happen in individual stocks all the time. That there are algorithms fighting algorithms all the time.  Its to the point where it sounds like the plot of a science fiction movie and it is.

Our markets are at risk.  I can’t quantify how big a risk there is, but I can tell you this, as long as there are algorithmically driven trades that happen in thousandths of a second, we can not eliminate that risk.  As long as that risk exists, there is a significant opportunity for hackers , terrorists (yes terrorists) and crooks to negatively impact our market to the tune of billions of dollars and possibly . This is the rise of the software controlled market. 

There is no such thing as bug free software.  If you are an investor, learn how to hedge.




31 thoughts on “The Real Problem with the Twitter HackCrash

  1. iTweet!
    My children will be rich.
    99 Domains!

    Comment by dylanisreddy -

  2. Pingback: Ted Nelson: Shinichi Mochizuki is Satoshi Nakamoto, the Inventor of Bitcoin and ABC Proof | Pink Iguana

  3. Pingback: A Hackcrash of Reality | The Weekend Muser

  4. Here are some real problems:

    • 30 million unemployed
    • American industry (steel, building materials, capital goods, textiles, consumer goods, etc.) all but disappeared due to free trade
    • Wall Street banks are nearly unregulated, sucked $26 trillion out of our national debt with the 2008 bailout, and have an open line of credit for $85 billion EVERY MONTH with QE3.
    • No economic recovery program proposed by anyone in the political mainstream.

    What to do?

    1. Stop programmed trading entirely with a 1% Wall Street Sales Tax. Protect household savers with a $1m exemption, get a sliver of the $5+ quadrillion each year of securities sold in NYC, Chicago and elsewhere.

    2. Force the Federal Reserve to act as a national bank to finance (interest free, with 100 year bond maturities) trillions in public infrastructure.

    3. Cancel the bailout and process the “too-big-to-fail” banks through an orderly liquidation process, with their worthless derivatives eliminated by triage.

    4. Reinstate Glass-Steagall, ban risky categories of derivatives.

    Want to read the rest of the program?

    Mr. Cuban, we’d love to have you aboard!

    Comment by deadeyeblog -

  5. Pingback: The Real Problem with the Twitter HackCrash

  6. Also, people that blame HFT for their trading failure or lack of commitment to learn the way of consistent trading are people with no self confidence and no persistence to put in the effort to do it. Quitters don’t win.

    Comment by Michael Samhan -

  7. Paul is right, there are a lot of studies going on right now to wire computers neurologically just like our brains are wired. They will be processing information faster than the majority of humans. Once this technology is very well implemented there is no telling how these things will be able to react to information. No one knows the future, but technology will not stay at its current phase. It will only get faster, smarter, and more accurate from this point forward.

    Comment by Michael Samhan -

  8. As information sources are shown to be prone to be unreliable (e.g., prone to hacking), markets will discount the information they provide. I can assure you the next time AP tweets about an event involving a terrorist attack on the White House that you won’t see the same impact on the markets.

    Comment by Doug Bailey -

  9. Why should the vast majority of investors (or humans) care about flash crashes? Unless you use stop-loss strategies in your investments, they should never have an effect on your portfolio’s value.

    Put another way, if you watch the market closely enough to care about and react to flash crashes, you (in the general sense, not just Mark) are a trader, not an investor. If that’s the case, your complaining about flash crashes is sour grapes, since someone else has found a better way of doing than you.

    Comment by Jeremy Filliben (@jfilliben) -

  10. I still don’t understand why an alternative stock market can’t be created in which logical rules apply, such as dividend requirements unless a company truly lost money, all stock purchases have to stay at least one month in the company before being resold, a super tiny transaction charge “tax” per order, I’d even like to see a “made in America” stock market in which all companies make their product in america.

    Comment by swarmthebanks -

  11. Algorithms of these machines are the reason why I no longer trade stock. I was never by any means a big investor, but when something else can do hundreds of trades in less than a second, I see it as an unfair advantage that can negatively impact me very quickly.

    Comment by Ron Stefanski (@RStefanski) -

  12. Pingback: Cybersecurity: Policy, practice and the future of hacking | The Balance Sheet

  13. Daywalker, I don’t think technology is being demonized. In fact, there are layers and layers of technology in even the MOST human interactions in the market these days.

    Certainly I’m not demonizing technology. I just understand and embrace the chaos that emerges from complex systems.

    The thing about tech is that it is, in many ways, the nature that man created.

    When you read a Jack London book, and some guy dies in a snow pile, I don’t think anyone demonizes the snow. It’s just a reality that if you are a human that goes out into the snow, you might die of hypothermia and you might get eaten by wolves. There is no judgment in that fact, simply presentation and silence.

    The same is true of technological systems. We set them in place, but when they go crazy, it is not the fault of the system. There is no malice in a piece of code, or in a reaction by a machine to a dataset.

    The fact is, machines can be tricked into doing terrible things, and unlike humans, do not have the capacity to reason or infer (unless that capacity has been designed into them, and even then, the A.I. is implemented by a human being who might fail to predict each case).

    Comment by Steven Kilpatrick -

  14. Algos can quickly read news and make trading decisions, but they can also quickly (within seconds) realize their mistake, like they did in this case. A human can make the same mistake, but in this case a human would probably react even slower and less efficiently. Algos (and technology in general) should not be demonized.

    Comment by Daywalker415 (@daywalker415) -

  15. With all the talk of people building new currencies like Bitcoin, we should actually go build a new *market*. Seriously.

    Comment by Chris Treadaway -

  16. Rather than being afraid of algorithms I’m looking to empower retail users with the same tools and level the playing field. We’re providing free financial data for engineers to design these algorithms and democratize/decentralize the strategy creation. We don’t work in HFT but longer term strategies.

    I really would love it if you were an advisor Mark. We just signed two major partnerships and seed funding (jared at

    Comment by Jared Broad -

  17. If there were a legion of Kurzweils designing and implementing those computer intelligence systems, @Paul Markgraff might be right. Alas, @Steven Kilpatrick knows firsthand that such systems still rely on the linear intelligence of humans.

    While that’s the case, heeding Cuban’s observation yields the best result: there’s a risk here and it’s getting worse.

    Comment by Danny Favela (@hyliandanny) -

  18. Early in my compsci education at Berkeley we engaged in software bot-wars which seem to match well the program trading of today. Of course, it never occurred to us to claim any social value for our actions.

    It appears, like our early experiments, an implicit goal of HFT is actually to NOT get caught providing liquidity to the market, (hence bids cancelled after a millisecond.)

    Time to outlaw it, and for me to go back to work. Great post Mark.

    Comment by Michael Farr -

  19. It’s important for everyone to realize that nothing happened in the markets for 22 seconds after the tweet. We talk of computers reacting in sub-millisecond timeframes, yet on this event, the algos took 22 seconds? Why did it take so long? Because this wasn’t a case of computers reading a tweet and reacting at all. The market started to move only after this tweet was announced on squawks, like Trade The News, and HUMANS reacted to the news they had just heard over the squawk.

    Comment by Josh Hansen (@joshua_hansen) -

  20. I think Stephen hit the nail on the head with his last sentence. Also, if a player is fooled by a fake jump shot will he not rethink his/her strategy next time? Algorithms and people evolve and learn. And who actually lost money in this case? the owner/seller of that algorithm or the buyer who discovered the anomaly?

    Comment by gatornation -

  21. There is a very simple fix to all of this — impose a tenth-of-a-cent tax on each share sold. Investors (as opposed to traders) aren’t moving hundreds of thousands of shares a day. For someone selling 500,000 shares of stock, that’s an additional $500 tax they’d have to pay, which is negligible in a transaction that size for an individual investor, but would greatly cool the insta-trade market that has been built around rapidly creating, then exploiting, imperfect information.

    And I’m guessing if you own half a million shares of any publicly traded stock, $500 is the change in your couch. Less than what you spend on lunch. It’s a tax that’s 100% progressive and solves an economy-threatening problem.

    Comment by speakhappiness -

  22. @Steven Kurzweil argues that as software/hardware systems surpass human intelligence, concepts such as “people attacking our systems in the public and personal space” will become less relevant. Of course, physical attacks are always a possibility. But to argue — as you do — that computer software will always be penetrable by human “hackers” demonstrates linear thinking instead of exponential/logarithmic expansion. Human intelligence improves on a linear scale. Computer intelligence improves on an exponential scale. In the near future, computer processors will outstrip human brain processing capabilities. It seems short-sighted of you, in this instance, to fail to recognize the inherent logical flaws in your argument.

    Comment by Paul Markgraff -

  23. We can’t get supposedly educated people to be able to discern between truth and nonsense in something as publicly provable as the existence of our president’s birth certificate! How are we ever going to think we can get machines to do it? Even if Paul is right and machine intelligence for recognizing deception “equals our own” within 10 years — that’s still not enough

    Comment by Mike Crowder -

  24. Terrific insight into this fast-evolving HFT world where the “arms race” has apparently just begun. It’s incidents like this that make us small investors feel wary about the stock market where the playing field appears to be more than a bit tilted! It reminds me a bit of the early days in the web when companies were paying “independent” bloggers to write glowing reviews of their goods and services — or to write diatribes about their competitors!!

    Comment by misterlapointe -


    Comment by Mücahit Coşkun -

  26. Paul, the notion that we can use the predictions in a book to predict the efficacy of the future is incredibly short sighted, despite the attempt at the opposite.

    I’m a game designer, and one of the things I can tell you is that systems are always exploitable. Often bugs happen accidentally as a result of unexpected combinations of cascading actions. Just as often, players work as hard as they can to find the exploits in your systems.

    Given that the people attacking our systems in the public and personal space are also becoming more sophisticated, it’s irresponsible to assume that a smart computer program is the same thing as an impenetrable one.

    Hell, the entire notion of “hacking” (the positive word for rooting around in code, not the negative term for subverting security) is about using a computer’s logic in ways not originally intended. The only thing you can say for sure about a more sophisticated piece of code, is that there will be even more sophisticated ways to subvert it.

    Comment by Steven Kilpatrick -

  27. Please take me off the list!

    Michael E. Douroux

    Comment by Michael E. Douroux -

  28. Mark, I’ve been a professional day trader for 12 years, and always value your perspective on the markets. Since around 2008, I’ve seen high frequency bots gradually kill a lot of the “old school” traders who can’t adapt. As much as I hate the thought of a purely computer driven market, we have to adapt to survive. That’s why I use a combination of automation with constant manual monitoring of the market conditions. I think anyone who plugs in a “set it and forget it” approach to trading is asking for BIG trouble.

    Comment by Chris Dunn (@ChrisDunnTV) -

  29. Mark, be sure to check out the email I sent you about an mma app. You’ll love it!!

    Comment by Pete Gustin -

  30. Thanks Mark! Very helpful info.

    Comment by lisarlee101 -

  31. “The problem is that no matter how smart you make machines, they will never be smart enough in our lifetime to detect all levels of deceit and fraud.”

    I would argue that this is incorrect. Refer to Ray Kurzweil’s book “The Singularity Is Near” to understand just how quickly machine intelligence is catching and will surpass our own. You’re looking at less than 10 years.

    Comment by Paul Markgraff -

Comments are closed.