Untitled record

0001The present invention relates to a method, a use and a system for warning a recipient of possible dangers by offenders.

0002Especially in recent times, mathematical models have made significant progress in modeling reality due to the ever-increasing amounts of data that can be collected and analyzed. Databases are increasingly linked and an attempt is made to predict future events using so-called big-data methods. However, today's technicality also has the disadvantage of helping criminal elements to network better and plan offenses more effectively. In particular, the burglary statistics in recent years show an increasing increase in burglary offenses. So, lately, people increasingly have an increased need to be warned about such burglary offenses or to be given an opportunity to prevent them in advance.

0003<patcit><text>US 2008/0238668 A1</text></patcit> shows an alarm system and method for operating this alarm system. Two building units are connected via a communication network. In each of the building units, an alarm system is mounted, with the respective alarm system is connected to other functional components via an intra-building network. In the event of an alarm, each of the alarm systems can selectively inform the other alarm system via the communication network of an alarm case and, if necessary, automatically activate the other functional components, so that residents of the houses are automatically informed. The disadvantage here is that the burglary can not be prevented for both building units in advance. It is not warned in advance, but the warning is coupled to a break-in, which takes place at this moment.

0004<patcit><text>US 2011/0046920 A1</text></patcit> shows a method for alerting a receiver carrying a mobile terminal, wherein the terminal determines position data and forwards via a communication network to a computer. On this computer an algorithm is implemented, which determines a possible threat situation for the receiver with the help of several data records and other parameters. The records include, for example, information about the area in which the recipient currently attracts attention or moves to, such as resident income, crime statistics, and so forth. The algorithm then determines the possible threat situation as a function of the data records and can automatically warn the recipient about his mobile terminal and possibly even organizations, such as the police, for example. notify. However, it is disadvantageous in the learned method not to warn an apartment against an acutely increased threat situation. Since the threat situation is determined by the algorithm on the basis of information about the eligible residential area, the algorithm can not respond flexibly to an increased threat situation within the residential area, since this does not change in the data set. In addition, the learned system in areas that are described in the record as "safe" would almost never detect an increased threat situation, even if offenders reside in this area. Since the threat situation is determined by the algorithm on the basis of information about the eligible residential area, the algorithm can not respond flexibly to an increased threat situation within the residential area, since this does not change in the data set. In addition, the learned system in areas that are described in the record as "safe" would almost never detect an increased threat situation, even if offenders reside in this area. Since the threat situation is determined by the algorithm on the basis of information about the eligible residential area, the algorithm can not respond flexibly to an increased threat situation within the residential area, since this does not change in the data set. In addition, the learned system in areas that are described in the record as "safe" would almost never detect an increased threat situation, even if offenders reside in this area.

0005Accordingly, it is the object of the invention to provide a method, a use and a system by which an increased threat situation in an environment is automatically determined and a receiver is warned of future offenses.

0006This object is achieved with the features of claims 1, 9 and 10th

0007According to the invention, the method for providing an automated warning to a recipient of possible offenses comprises the following steps:<ul list-style="bullet" id="ul_0001"><li id="ul_0001_0001">• Create a first set of offense information on a server, each of which has a timestamp and location information indicating where and when the offense occurred.</li><li id="ul_0001_0002">• Creating a second data set, in particular by a network provider, wherein the second record has unique identification data from mobile devices and a time and location.</li><li id="ul_0001_0003">• Merge the first and second records into a population of data on the server.</li><li id="ul_0001_0004">• Defining temporal and local limits of residence with respect to individual offenses, whereby the temporal limit of stay can be described by a period of time and the local limit of residence by a radius around the time and place of the offense.</li><li id="ul_0001_0005">• Identify mobile devices that are within the limits of the offense.</li><li id="ul_0001_0006">• Identify and mark the mobile devices that can be found in several residence limits.</li><li id="ul_0001_0007">• Automated alerting of the recipient when one of the tagged mobile devices is within a recipient-defined environment.</li></ul>

0008The first record can be provided by authorities such as the police and / or alarm systems. It is also possible for users to display a crime via a mobile application on their smartphone, which is then sent to the server and added to the first record. The data of the second data set can be extracted by the network provider during the radio cell change of the mobile terminals when the mobile terminal is connected to another transmitting station in order to ensure optimal reception. In this case, the network provider can access the necessary data, since it is known when the device connects to which transmitting station. Also a time-resolved GPS-location is possible. The unique identification data of the mobile terminal can, for example, via the SIM card, other subscriber identity modules, IMEI and / or unique identifier data may be obtained from apps such as Facebook or Google Maps. The definition of temporal and local limits of residence for individual offenses can be stated as follows: a first offense is assigned a time span of half an hour at the time of the offense and a radius of 2 km around the place of the offense as the limit of residence, which means a mobile terminal located within 15 minutes of the time of the offense 500 meters from the location of the offense is identified as being within the limits of the stay. The residence limits can be set individually for each offense. The automated alert of the recipient is triggered by the server, wherein the automated alert is preferably sent as a message to a mobile terminal of the recipient. The message can also be sent to a computer and / or the police will be notified immediately.

0009The invention thus advantageously makes it possible to identify mobile terminals, in particular smartphones, which are frequently in the vicinity of offenses relating to both spatial and temporal proximity. For example, individuals or groups of perpetrators can be identified who have committed offenses with an increased likelihood in the past and therefore could commit offenses again, since it can be assumed that they carry their smartphones mostly with them. This method is particularly effective in automatically warning the recipient against thieves that they usually "clump" around a crime. For example, a group of police officers may be distinguished from a group of thieves in this process, that the mobile terminals of the police are usually only "after" the time of the offense within the boundaries of resident. The method thus advantageously enables the recipient to be reliably informed of an increased threat situation and to be able to react accordingly.

0010Preferably, the automated warning takes place only after exceeding a threshold, which determines the probability that the carrier of the mobile terminal is liable to a risk of delict. One possible choice of the threshold is that the automated alert is triggered only when a mobile terminal of a certain number, for example, five times, was found within the limits of stay. Another possibility is to set the threshold using statistical methods of hypothesis testing as a significance value, eg alpha = 0.05. This has the advantage that the sensitivity of the automated warning can be adjusted, since otherwise would be warned of all carriers of the mobile terminals, which were even once within the residence limits of a crime.

0011Preferably, the receiver can set the threshold himself and thereby affect the sensitivity of the automated alert according to his wishes and needs. For example, the owner of a jewelry store, who has an increased risk of becoming a victim of a crime, may choose to be more sensitive.

0012In a preferred embodiment of the invention, an algorithm is implemented on the server, which determines correlations between the crime information and the mobile terminals from the data aggregate and creates a predictive model, this predictive model being based on statistical methods. The advantage of a complex statistical model over a simple frequency analysis is that additional parameters can be used in such a predictive model that further increase the predictive power. Other parameters that can be included in the predictive model to better model the risk for a particular neighborhood are median income, highway distance, development structure, age of the development, age of residents, state, distance to the nearest police station, and / or other parameters. Here, the algorithm is designed as a trainable algorithm in the form of a "machine learning algorithm" whose predictive power is steadily improving. The more fully the mobile devices as well as the offenses are detected, the better the predictive model. The algorithm may be implemented as a neural network or as a decision tree algorithm. In addition, clustering methods can be used.

0013Clustering methods or cluster analyzes are methods for the discovery of similarity structures in large databases. In the present case, for example, the identification of the mobile terminals which have been identified several times within the limits of stay. The groups of similar objects found in this way are called clusters and the group assignment is called clustering. In cluster analysis, the goal is to identify new groups in the data - as opposed to the classification where data is mapped to existing classes. The numerous algorithms that are used in cluster analysis differ mainly in their similarity and group concept, their cluster model, their algorithmic approach and tolerance to disturbances in the data.

0014Preferably, the predictive model calculates a probability value for the occurrence of a crime that is compared to the threshold. Ultimately, it is a matter of defining if the automated alert is triggered if the probability value falls below or exceeds the threshold value. It is important to note that ultimately a certain probability is to be ensured with which a crime takes place. The statistic usually uses a significance level of alpha = 0.05, which means that the hypothesis that a crime occurs in 95% of the cases would be correct for a random sample. One statistical way of qualifying this is, for example, the so-called p-value of the chi-square test.

0015This will be explained by means of an example. The mobile terminals A, B, C are grouped into a cluster ABC and are more often in the vicinity of a crime to identify than the mobile terminals D, E, F, which are combined into the cluster DEF. To illustrate the following four-field tables: In the case of the ABC cluster, one offense was committed 19 times and no offense six times if the corresponding mobile devices were within a specified radius at the same time.<tables num="0000"><table frame="all"><tgroup cols="3" colsep="1" rowsep="0"><colspec colname="col1" colsep="1" colwidth="34*" /><colspec colname="col2" colsep="1" colwidth="55*" /><colspec colname="col3" colsep="1" colwidth="36*" /><thead><row rowsep="0"><entry align="left" colname="col1" valign="top" /><entry align="center" colname="col2" valign="top">"Offense"</entry><entry align="center" colname="col3" valign="top">"No offense"</entry></row></thead><tbody><row rowsep="0"><entry align="left" colname="col1" valign="top">Cluster ABC</entry><entry align="center" colname="col2" valign="top">19</entry><entry align="center" colname="col3" valign="top">6</entry></row><row rowsep="1"><entry align="left" colname="col1" valign="top">Cluster DEF</entry><entry align="center" colname="col2" valign="top">2</entry><entry align="center" colname="col3" valign="top">20</entry></row></tbody></tgroup></table></tables>

0016In this case, the Chi-Square test for the significance level of 0.05 gives a p-value of 0.00004. This means that the original hypothesis that there is no difference between clusters ABC and DEF must be discarded. The inference could be that cluster ABC represents a thief group, causing an automatic alert if the mobile terminals of that cluster are within a given radius.

0017There is also the possibility that customers are informed of different warning levels on a scale to be determined arbitrarily. Furthermore, the data can be displayed on digital maps in real time. In addition, historicized probabilities can be mapped or aggregated values ​​can be predicted locally for any area.

0018The server can also automatically inform a recipient's smart home system of the increased risk condition so that an alarm system connected to the smart home system can selectively increase its security level. So camera and image recordings can be started or the lighting can be turned on in homes to deter thieves or burglars.

0019According to another possible embodiment of the invention, the offense information and / or the unique identification data are again removable from the data aggregate. This is advantageous, for example, when it has become known that the mobile terminal has permanently changed its owner. The new owner / carrier can not and should not necessarily be related to past offenses with which he had nothing to do. Typically, mobile terminals are also very short-lived, so storing all of these records would require an unnecessarily large volume of data since the mobile terminals are typically used for no more than five years. It can thus be provided, for example, that the algorithm contains delict information or Identification data automatically removed from the databases after five years or stored in compressed form. Also, the algorithm should weight more recent identification data and offending information in its analysis.

0020In accordance with another aspect of the invention, there is provided the use of the above-described method for automatically identifying potential offenders. By being able to assign an owner or carrier to each smartphone normally, it is possible to use the method to identify and monitor potential perpetrators, including potential terrorists. If it then turns out that the suspicion was unfounded, the corresponding smartphone can be deleted from the dataset.

0021According to another aspect of the invention, there is provided a system for performing the method, the system having a server on which the algorithm is implemented for evaluation and on which the first data set is available. In addition, the system has a communication network for transmitting the second data set and terminals, in particular smartphones, wherein the terminals automatically transmit the second record to the server by means of the communication network. Here, the algorithm analyzes correlations in the data entity and, in the event of a potential hazard via the factory communication network, sends a warning to a user's device capable of receiving messages.

0022In a further advantageous embodiment of the invention, all collected data are pseudonymized directly, ie each date (eg IMEI) is converted via a secure instance into an unrecoverable date. For this purpose, algorithms and / or mapping tables are used. This ensures that the customer data used can not be misused and customers are not compressed. Exceptions may, however, be made by the respective legislator.

0023Further advantages, features and development of the invention will become apparent from the independent claims and from the following description of preferred embodiments.

0024The invention will be explained in more detail below with reference to the accompanying drawings with reference to preferred embodiments.

0025Show it<ul list-style="none" id="ul_0002"><li id="ul_0002_0001"><figref>1</figref>: an overview of a radio cell area with a mobile terminal in one of the radio cells.</li><li id="ul_0002_0002"><figref>2</figref>: the overview <figref>1</figref> and an additional receiver.</li><li id="ul_0002_0003"><figref>3</figref>: a schematic overview of the data analysis of the method according to the invention.</li></ul>

0026<figref>1</figref> shows an overview of a radio cell area <b>5</b>showing a plurality of radio cells having a hexagonal structure. In<figref>1</figref> centrally located is a home radio cell <b>10</b>from neighbor radio cells <b>12</b> is surrounded. Inside the home radio cell<b>10</b> are an apartment <b>11</b> and a mobile station <b>16</b> located. The hexagonal structure of the home radio cell<b>10</b> can be used to set a spatial limit <b>17</b> define. Also within the home radio cell<b>10</b> there is a mobile device <b>14</b>, in particular a smartphone, which carries an owner and that with the mobile radio transmitter <b>16</b> is connected and optionally has a detection application. <figref>2</figref> essentially corresponds <figref>1</figref>, being instead of the apartment <b>11</b> now a receiver <b>20</b> with his mobile device <b>22</b>, especially a smartphone <b>22</b>, in the home radio cell <b>10</b> located.

0027<figref>3</figref> shows a schematic overview of the data analysis of the method according to the invention. A created first record<b>30a</b> with offense information and a created second record <b>30b</b> with unique identification data of the mobile terminals <b>14</b> become a data aggregate <b>34</b> merged and in the next step by an algorithm <b>36</b> evaluated. The algorithm<b>36</b> creates a first predictive model <b>38a</b>as an input parameter for re-analysis by the algorithm <b>36</b> serves. This re-analysis will be followed by another first record<b>30b</b> and another second record <b>32b</b> added. As a result, the algorithm provides<b>36</b> a now trained predictive model <b>38b</b> of higher quality. Represents the trained predictive model<b>38b</b> Determining that a threat is likely, it transmits this information to a web service <b>40</b>who has a communication network <b>42</b> a warning to a smartphone <b>22</b>, a smartwatch <b>24</b> and / or a PC <b>23</b> and / or to another device of the recipient <b>20</b> sends.

0028The method is explained below with reference to two exemplary embodiments.

0029Example 1: The unique identification data of smartphones <b>14</b> are time-resolved for each radio cell used 10,12 and pseudonymized in a big-data solution (eg Hadoop platform) as a second record <b>32a</b>, b saved. For all radio cells 10,12 exists a database which has information about the geographical area of ​​the radio cell 10,12 located in it. In the first record<b>30a</b>, b are recorded locally and temporally resolved offenses. The algorithm<b>36</b> determined based on the located in the radio cells 10,12 smartphones <b>14</b> and other local parameters over the geographical area (such as median income, age structure, distance to the highway, ...) a prediction of the local burglary risk. For example, the result may be displayed on a digital map and / or if the threshold is exceeded, an automated alert will be sent to the recipient<b>20</b> Posted.

0030The algorithm <b>36</b> puts the smartphones <b>14</b> in temporal and spatial correlation to the offenses, by an evaluation of the data set <b>34</b>, Here, both time and location information can be used as parameters for the predictive model<b>38</b> be used. Consider the following situation:<ul list-style="bullet" id="ul_0003"><li id="ul_0003_0001">• Smartphone A was within the home wireless cell for 13 minutes 49 minutes before a break-in <b>10</b> of the crime scene.</li><li id="ul_0003_0002">• The smartphones B and C were 1 hour before the break-in in the neighboring radio cell <b>12</b> localized for a period of 2 hours and at the time of burglary within the home radio cell <b>10</b> localized.</li></ul>If such or similar relationships with other offenses and the same smartphones A, B, C occur, they can be recognized and counted by clustering. The smartphone A falls into the cluster "potential scouts" and the smartphones B and C into the cluster "potential burglars". In addition, due to the counted correlation with offenses, a weighting coefficient can be determined which is used for later training of the predictive model<b>38a</b> is being used. The predictive model<b>38a</b> is additionally trained with other available data (such as the development, median income, burglary rate per inhabitant, age of the building, inhabitant, time, date, etc.). For this purpose, a first training group is formed, were present at the offenses and it is formed a second training group in which there were no offenses. Both groups should be chosen for statistical reasons so that they are about the same size. The result is the predictive model<b>38a</b> trained on the answer "burglary" or "no burglary". As a result, a trained predictive model<b>38b</b> generated that can be tested with other available data in terms of its quality.

0031If the smartphones A, B and C now meet the cluster conditions with real-time data, they will be replaced by the predictive model <b>38b</b> and calculates a local probability of another burglary. If a threshold to be defined is exceeded, a stepped alarm is issued via the communication network<b>42</b> to the recipient <b>20</b> or a customer.

0032For the predictive model <b>38</b> The clusters can be formed with the data in such a way that larger clusters can be formed and models can be determined for certain burglary types. Such clusters can be, for example, "window break-ins", "burglaries in rural areas" or "serial burglaries". These clusters can also be formed by combining the various parameters. The predictive models<b>38a</b>, b will then be for all smartphones <b>14</b> calculated that fall into these clusters. The algorithm<b>36</b> learns from the behavior of several potential burglars who have similarities and thus fall into a cluster. To select the best predictive model<b>38a</b>, b can also be automated and evaluated several methods or algorithms and compared. The evaluation takes place on the basis of historicized data from the past. Alternatively, the clustering can be dispensed with and the existing data is directly incorporated into the predictive model<b>38a</b>, b used.

0033Example 2: Here, the prediction model becomes <b>38a</b>, b is not trained on the basis of burglary data, but the first record <b>30a</b> is made up of data from the receiver <b>20</b> even by means of his smartphone <b>22</b> recorded and made available. This can be done for example by a special application in which sensations can be detected that are related to surrounding people. Examples include "feel threatened", "feel uncomfortable", "am feeling good" or "might fall in love". Additional data included in the first record<b>30a</b> Information about certain places (near the train station, shopping center, ...) or events (football match, ...) could be received. Also in this case, existing offenses such as handbag theft or robberies can be detected. The corresponding predictive model<b>38a</b>, b is trained in the same way as described above. However, here are only present at the same time smartphones<b>14</b> Considering that persons present in the past have no influence on the interactions of the customer / recipient <b>20</b> to have. This allows the predictive model<b>38a</b>, b be trained to the receiver <b>20</b> or send another message or warning to another customer.

0034Further modifications within the scope of this invention are that the method described can be used to identify suspicious persons and report them to the police, for example. The police can be provided with a special analysis view for the early detection of dangers. Security partners of a customer may receive relevant alerts and the method may also be used to send advertising information tailored to a customer.

QUOTES INCLUDE IN THE DESCRIPTION

0000This list of the documents listed by the applicant has been generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Cited patent literature

0000<ul compact="compact" list-style="dash"><li>US 2008/0238668 A1 [0003]</li><li>US 2011/0046920 A1 [0004]</li></ul>