Broker timing attack possible?

New comment
 

In looking at historical charts downloaded from various demo accounts, once I take the time to determine "clock skew" between the data and align it for side-by-side comparisons, I've noticed that there is a significant difference between brokers in prices. The differentials well exceed the pip spread, such as on Feb 3, 2006 at (roughly; don't know who has the exact right time) 14:29 CET I show a significant drop (-30 pips) in GBPUSD on the M1 data from some brokers like Alpari, but from other brokers I see a sharp rise of 30 to 40 pips. Indeed I could chalk this up to basic skew, bad data, etc... however the rise is held for over 5 minutes with these other brokers.

Because I am very new to forex, I am not entirely sure what to make of this. The paranoid will say it's just basic stoprunning to deal with poor hedging on the broker's side. However, my question is this: Can we not grab various tick data from many brokers, reconcile it all to a single time source (say a simple n-way commit database), and look for patterns? (DDE seems to be a likely technique for scraping the data... preferably from multiple points world-wide along with ping samples every minute to measure network latencies, etc, etc)

I have a theory that preceeding many larger moves, we can get very solid hints from data mismatch like the example I give above. Think about it, one broker goes -30, the other goes +30. That's 60 pips arbitrage right there. Of course, we cannot really benefit from arbitrage between brokers unless we can move serious volume... but can we predict (yes, Predict) the future price movement by "stealing" the price data that brokers hold from us while they cover their own asses?

I think this type of pattern would exist even with commercial data streams, since they merely get thier quotes from specific banks within the interbank...

I call this a timing attack because it's exactly what crackers do to crack smart cards, encryption algos, etc... of course, these devices are mathematically proven to be secure, so timing attacks are really just statistical guesses. Brokers, seem to be, street folk who play gambling games... maybe just CFA-level knowledge, if you're lucky. Seems easy to 'exploit' this 'vunerability', even with slippage and other filling tricks on their side.

Anyhow, as I said, I'm really new to Forex so feel free to flog me if this is crazy talk. I post when I lose... gives me "control"

PS: And yes, I know this drop was because of news.

 

Nice theory. There are many members have multi-accounts of brokers. They may have something to share

 

I had similar stuff in mind some time ago. to go ahead on this idea, you first have to do a systemic research first coving data for some years I think. you also has to consider the clearing cost if you don't do the opposite transactions at each broker. if you do the opposite transactions at each broker, then the price diffs must be larger than the sum of the two brokers spreads. considering slipage,the price diffs maybe must larger than twice of that. then you need some programing skill,or maybe somebody with programming skill will cooperate with you.

 
suffic369:
I had similar stuff in mind some time ago. to go ahead on this idea, you first have to do a systemic research first coving data for some years I think. you also has to consider the clearing cost if you don't do the opposite transactions at each broker. if you do the opposite transactions at each broker, then the price diffs must be larger than the sum of the two brokers spreads. considering slipage,the price diffs maybe must larger than twice of that. then you need some programing skill,or maybe somebody with programming skill will cooperate with you.

I wasn't thinking to exploit the spread between the brokers necessarily.

I believe there may exist a pattern as to who leans which way first to signal an underlying legitimate price move that we just can't see yet. I assume brokers have access to price data well in advance of release, especially if they have their own hedging to secure prior to allowing the move to happen. What surprised me is the distortion can appear to happen as long as 5 minutes!

As you say, live collection of data over a long period may be necessary. I do not trust the historical data fed from these places, however who knows. I'm sure they don't waste time to sanitize the history, lest risk actual clients coming back to say "Hey, that wasn't the price I saw 3 days ago!"

 

I think the point in time you mentioned was right before or at the time of the release of the NFP data. So there might be big differencees for a couple of seconds of fractions of seconds there. But it would be very difficult and risky to try arbitrage during the NFP release because it's not sure you would get your order filled at this special time.

And if you only get one side filled you have a problem.

I thought about this technique too and i think it's interesting but also risky.

Edit: You saw a difference for 5 minutes?? That's very special, i never saw this. Wow.

 

I have been noticing this kind of behavior especially with gold. I've been watching the feed from 0DL and they seem to have the most fluid price changes and a lot of the other firms will follow this price a few seconds (to a few minutes) later. The long time period is usually with M1G, however they will constantly change their prices whenever you try to buy at the lagged rate (has happened a half-dozen times where the price will remain unchanged for minutes and the second you send the order they re-calculate the current price - not cool).

Watching one broker feed skyrocket in price while the other stays still for a few seconds can give you a foot in the door but of course could backfire so be careful with it.

Does anyone know where the original feeds come from or who has the best chance of getting the most current market price?

New comment
Reason: