Methods and systems for reducing the spread of files on a network
View Patent ↗Methods and systems can reduce the spread of computer files or data on a network by obtaining and tracking times of arrival for chunks of data transmitted on the network. The times of arrival for a node can be transformed into time-series and periodograms computed from the time-series. Successive periodograms can be compared to determine changes in the strongest peaks of the periodograms. If a new peak is identified, a search for the occurrence of the peak in previous periodograms can be conducted. If no peak having a matching frequency is found, a search for the peak in the periodograms for neighboring nodes can be performed. If matching peaks are found, the associated data stream can be classified. Predictions of the timing and length of associated data packets can be used to randomly interrupt transmission of associated data packets resulting in reducing the spread of the classified data stream.
1. A method of classifying disruptive data packet traffic flow on a network, comprising:
obtaining at a router, periodograms based on times of arrival of said data packets from nodes of said network;
comparing successive periodograms for each of said nodes to determine when at least one new peak above a threshold is present in one of said periodograms for one of said nodes;
marking said at least one new peak as a suspicious peak when said at least one new peak does not occur in one of a predetermined number of previous periodograms for said one of said nodes;
determining if said suspicious peak occurs in at least one of a specified number of periodograms obtained at said router for others of said nodes of said network;
determining if said suspicious peak occurs in at least one of said specified number of periodograms obtained at other routers for said one of said nodes; and
providing a signal to classify traffic flow corresponding to said suspicious peak as disruptive traffic flow when said suspicious peak occurs at least in at least one of said specified number of periodograms obtained at said router for said others of said nodes and in at least one of said specified number of periodograms obtained at said other routers for said one of said nodes.
2. The method of claim 1 , wherein obtaining said periodograms further comprises:
tracking times of arrival data;
transforming said times of arrival data into time-series;
parsing said time-series into windows; and
obtaining Lomb periodograms for said windows.
3. The method of claim 1 , wherein providing a signal to classify traffic flow uses a Hidden Markov Model.
4. The method of claim 1 , further comprising:
interrupting data packets by randomly removing data packets for a specified time;
determining, after said specified time, if said at least one peak at said frequency occurs in at least one additional periodogram based on times of arrival of data packets since beginning said interrupting; and
determining, based on including said additional periodograms with said specified number of periodograms, when said at least one peak at said frequency occurs in said at least one additional periodogram.
5. A non-transitory computer-readable medium containing instructions for controlling a processor to classify disruptive data packet traffic flow, by:
obtaining at a router, periodograms based on times of arrival of said data packets from nodes of said network;
comparing successive periodograms for each of said nodes to determine when at least one new peak above a threshold is present in one of said periodograms for one of said nodes;
marking said at least one new peak as a suspicious peak when said at least one new peak does not occur in one of a predetermined number of previous periodograms for said one of said nodes;
determining if said suspicious peak occurs in at least one of a specified number of periodograms obtained at said router for others of said nodes of said network;
determining if said suspicious peak occurs in at least one of said specified number of periodograms obtained at other routers for said one of said nodes; and
providing a signal to classify traffic flow corresponding to said suspicious peak as disruptive traffic flow when said suspicious peak occurs at least in at least one of said specified number of periodograms obtained at said router for said others of said nodes and in at least one of said specified number of periodograms obtained at said other routers for said one of said nodes.
6. The non-transitory computer-readable medium of claim 5 , further containing instructions for controlling said processor to obtain said periodograms, by:
tracking times of arrival data;
transforming said times of arrival data into time-series;
parsing said time-series into windows; and
obtaining Lomb periodograms for said windows.
7. The non-transitory computer-readable medium of claim 5 , further containing instructions for controlling said processor to provide a signal to classify traffic flow using a Hidden Markov Model.
8. The non-transitory computer-readable medium of claim 5 , further containing instructions for controlling said processor to interrupt, by:
randomly removing data packets for a specified time;
determining, after said specified time, if said at least one peak at said frequency occurs in at least one additional periodogram based on times of arrival of data packets since beginning said interrupting; and
returning to estimating based on including said additional periodograms with said specified number of periodograms when said at least one peak at said frequency occurs in said at least one additional periodogram.
9. An apparatus to classify disruptive data packet traffic flow on a network, comprising:
a processor;
a memory in communication with the processor and containing program instructions, wherein the processor executes program instructions contained in the memory and the program instructions comprise:
obtain at a router, periodograms based on times of arrival of said data packets from nodes of said network;
compare successive periodograms for each of said nodes to determine when at least one new peak above a threshold is present in one of said periodograms for one of said nodes;
mark said at least one new peak as a suspicious peak when said at least one new peak does not occur in one of a predetermined number of previous periodograms for said one of said nodes;
determine if said suspicious peak occurs in at least one of a specified number of periodograms obtained at said router for others of said nodes of said network;
determine if said suspicious peak occurs in at least one of said specified number of periodograms obtained at other routers for said one of said nodes; and
provide a signal to classify traffic flow corresponding to said suspicious peak as disruptive traffic flow when said suspicious peak occurs at least in at least one of said specified number of periodograms obtained at said router for said others of said nodes and in at least one of said specified number of periodograms obtained at said other routers for said one of said nodes.
10. The apparatus of claim 9 , wherein obtain said periodograms further comprises:
track times of arrival data;
transform said times of arrival data into time-series;
parse said time-series into windows; and
obtain Lomb periodograms for said windows.
11. The apparatus of claim 9 , wherein provide a signal to classify traffic flow uses a Hidden Markov Model.
12. The apparatus of claim 9 , further comprising:
interrupt data packets by randomly removing data packets for a specified time;
determine, after said specified time, if said at least one peak at said frequency occurs in at least one additional periodogram based on times of arrival of data packets since beginning said interrupting; and
determine, based on including said additional periodograms with said specified number of periodograms, when said at least one peak at said frequency occurs in said at least one additional periodogram.