DPS DEPLOY (Department of Public Safety Deploy) is a real time decision support tool for improving the security of the University of Southern California and adjacent community. The tool can potentially be applied to strategic security forecasting (issues and required resources) for future public safety planning and deployment. The system anticipates adversary activities, assesses the risk of criminal and potential terrorism events on the USC University Park Campus and surrounding areas and recommends security resource allocations to address the risks. This system is focused on risk-driven data analysis and decision making.

Brief Overview­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­

The initial prototype (DPS_DEPLOY 0.5) of the DPS Deploy system is in the “Prototype Design” phase of the CREATE Research Transition Pipeline. It has been implemented by adapting and extending the InfraSec risk management technology used in seaport security. The prototype is implemented in Java and is portable to various platforms. The objective is to transition, productize and introduce DPS Deploy technology to a wide range of government and private sector markets. It is anticipated that CREATE would monetize the underlying intellectual property and Goodwill, resulting in CREATE receiving royalties on the products developed and the government and private sectors lowering their costs of providing security.

Current DPS_DEPLOY Capabilities

The DPS_DEPLOY 0.5 system takes as input historical (2011-2014) DPS crime data and generates a set of customized heat maps. Each heat map geospatially displays color coded regions within the USC campus and surrounding areas. Colors represent risk levels ranging from green (low risk) to red (high risk), gray regions indicate the absence of data. The generated heat maps allow for quick assessment of risk based on selected options:


Risk type options: display risk heat map per crime type: infractions, misdemeanors and felonies. User can select, one, two, or all crime types.

Time options:

  • All historical data
  • Specific year
  • Semester (fall, spring, summer)
  • Specific date
  • Specific time of the day
  • Week-ends/week days

The prototype supports a combination of some of the time-related options, for example selecting specific year (e.g., 2012) and specific time of the day (e.g., 10am) will generate a heat map based on the crimes committed at 10am during the year of 2012.

Once the risk and type of time options are selected, the prototype generates a risk heat map that represents the selected crime types during the specified time period.

To obtain additional details regarding each region, a user may select the desired region and that region’s related pop-up will appear. To drill down into further details of each crime (crime code, reporting officer, etc.), the user simply selects the relevant tab on the display.

In addition to the customized risk heat maps, the prototype displays limited data analytics. In particular, it illustrates the percentage of risk per region type, per crime type, as compared to the total campus risk (total number of regions) for the crime type. Currently defined region types are: open space, dining facilities, residence halls, libraries, parking, galleries/lectures/performances, parks and sports fields. For example, parking regions contribute 20% of felonies of the total felonies committed in all region types during the year 2013.

In addition to displaying customized risk hit maps and data analytics, the prototype supports assessment of risk dynamics based on the countermeasure allocations. Currently, supported countermeasures are: patrol cars, patrol bicycles, patrol T-3s (3-wheel scooter), foot patrols, and Contemporary Services Corporation (CSC – a contract security provider) patrols. A user can “drag and drop” available countermeasures to a selected region and observe the associated change in risk level related to the region (and possibly other regions).

The current risk model employed by the prototype is relatively simple: risk is a weighted sum (per crime type) of number of crimes per region per selected time interval.