Targeted Crackdown on Tax-related Crimes in Wuhan

Weng Xiao Dong

The Economic Crime Investigation Division of the Wuhan Public Security Bureau （hereinafter referred to as the WHECID） concluded two operations in a nationwide campaign targeting tax-related crimes in November 2019 and August 2020 respectively. The cracked cases totaled 6.71 billion yuan， of which 1.07 billion were tax dues. The divisions successes is attributable to an upgraded guiding philosophy that highlights the use of data and intelligence in real practice.

Tailoring investigative approaches toward the characteristics of tax-related crimes， the WHECID has adopted specialized analytical tools to assess different types of data， using algorithms to discover patterns in suspicious behavior. By doing so， the division has upgraded its investigation model from passive response toward a preemptive， active and scientific system， notably improving its capacity to fight tax-related crimes.

For a better understanding of Wuhans experience in combating tax-related crimes， we are delighted to have Zhang Ming， Head of the WHECID， to elaborate on their practices.

Mr. Zhang， Could you please firstly provide us with an overview of the WHECIDs performance in combating tax-related cases in recent years？

We have been constantly improving our case management capacity and more fully tapping our potential to deal with tax-related crimes since 2016. Weve adopted different analytical approaches and data analysis models to tackle different types of crimes， such as the large-scale issuing of fake or inflated invoices by fraudulent shell companies， the issuing of inflated general VAT invoices， and the industrial use of inflated invoices. Such practices were addressed in the “1208” Exemplary Case in 2016， the “331” Exemplary Case in 2019， and the “Lunjian 2018” and the “Lundao 2019” Nationwide Economic Crime Investigation Campaigns， to name a few. These efforts made us more capable of tackling tax-related crimes in multiple dimensions and greatly enriched our strategic thinking.

Can you elaborate with examples？

On June 12， 2020， the WHECID received a national “Cloud Operation” notice from the Economic Crime Investigation Division of the Hubei Public Security Department to collect evidence for the “12·04” case concerning inflated invoices issued by a suspect surnamed Shao and his accomplices， originally filed by the Economic Crime Investigation Division of Linhai City， Taizhou， Zhejiang? province. Involving over 20 provinces and cities the case deserves a full-chain strike， hence the nationwide operation.

To handle this case， the WHECID redoubled its efforts， utilizing information technology and big data， coordinating with various departments， and initiating a platform for cooperation between the police and tax administrations. After three days and nights of continuous investigation， we found that a suspect surnamed Chen and his accomplices took advantage of the contact-free delivery of general VAT invoices during the Covid-19 epidemic and issued fraudulent general VAT invoices in the names of shell firms set up with fake identities.

On June 19， the WHECID， guided by the instruction of the Ministry of Public Security， successively dismantled four sites issuing inflated invoices and collected such devices for criminal purpose as business licenses， seals， tax printer， disks， and mobile phones. Preliminary investigations showed that this criminal clique established 89 shell corporations between April and June 2020 and issued fake general VAT invoices with total par value of 529 million yuan， causing a 9.58 million yuan loss in tax revenues.

What experience has the WHECID gained from these cases？

It is paramount to adopt innovative concepts and creative thinking. In order to improve our capacity to tackle tax-related crimes， the WHECID follows through on new ideas and approaches， and takes preemptive measures tailored to different industries. We closely analyze both the macroscopic situation and microscopic case details to progress in our work.

Firstly， we must move from a passive approach to an active one. As the Internet and modern information technologies develop at breakneck speed， economic crime investigation authorities and tax administrations are ever more determined in combating fraudulent shell companies whose crimes constantly evolve too. In the past， inflated invoice cases were committed by large-scale shell firms in simple methods， while now they are perpetrated by smaller but better organized criminal cliques. The crimes tend to be disguised in diverse forms and committed through complex means with application of high and complex technologies， and usually via virtual platforms.

In the past， the WHECID could only analyze existing cases for solutions. Due to limited time， the understaffed WHECID was very passive in responding to merely a few cases. But now we have improved our strategic thinking and established new data analysis models which adapt to various case types and changing modus operandi regardless of the time frame. In this light， we can closely adapt to new policy orientations and market developments， take initiative in discovering new modus operandi， and uncover more solid evidence for data analysis.

Secondly， we have adopted tailored methods targeting crimes in different industries. Years of experience in dealing with tax-related crimes tells us that inflated invoice offenses among legitimate entities tend to become industry oriented. In other words， complete chains have been established in each of these industries for the purpose of committing such crimes. Therefore， its of great significance to study each crime model in different industrial scenarios. Addressing inflated invoice crimes in various industries has been a longtime task for the WHECID. In the past few years， weve closed cases in medicine， sheet materials， gold， and other industries， and summarized the features of abnormal behavior models in the aforementioned fields.

According to investigation， the WHECID has found that inflated invoice crimes are more likely to take place in commodity trading activities. By running analysis on the features of these commodities and tax notes， and the financial data of previous cases， we figure out that such commodities， including mobile phones， computers， electric wires and cables， all fall into the categories of bulk commodities and consumer products， sharing features across electrical， wire and cable， construction materials， coal， daily-use chemical， food and apparel industries. With solid evidence， the WHECID has made real progress in targeted crackdown on such crimes.

Lastly， it is important to integrate the holistic approach and the tailored approach. Tax-related crimes have their own special characteristics. Since every company has to deal with tax-related issues， the tax system is related to local companies in various ways. The most typical type of data found in tax-related crimes is tax note flow， which contains many key factors such as tax-payers ID numbers， time， price with tax included， commodity description， sellers and buyers， and legal persons.

Given these factors， we can start an investigation by analyzing tax note flow in an effort to figure out a companys general situation： its major products， business scope， competitors， the industry it belongs to， and operation condition. We can also draw a bigger picture of the whole industry： its features， share of tax revenue in the city， profitability and stability， as well as the economic mix of the city its based in. All these will provide technical context for us to keep track of the general situation of the industry and redress operational problems in a timely manner.

Are there any innovative measures in exploiting intelligence and information technology？

The WHECID remains intense in cracking down on tax-related cases. We have introduced a full range of technologies to raise our ability to tackle such crimes with stronger initiative， higher accuracy， and greater intensity.

We have developed the Prism system based on the characteristics of shell companies. It helps to identify shell company clusters by analyzing company registration information. To put it simply， we divide the process of case-handling into three sub-processes， i.e. spotting a suspicious shell firm， narrowing down criminal groups and fighting downstream crimes. We have developed models based on business registration information. In this way， we are able to analyze such crimes with a powerful combination of both traditional and technological investigation methods.

In the meantime， weve also developed the “PM” shell firm screening model on the basis of years of experience and the features of shell firms. It helps us to identify suspicious entities among millions of companies. Thanks to a cooperative platform between the police and tax administrations， we can retrieve taxpayers information， go through data filtration and classification， and use such data in combination or separately to identify and combat criminal groups. Weve upgraded the PM model from rule-based algorithms to decision tree algorithms， which assist to screen out suspicious firms. Then we can launch precision strikes against the weakest links of criminal groups based on specific rules.

Capital data reflect the characters of criminal gangs and the role of each accomplice. Therefore， weve developed the “Xingtu” all-transactions data analysis system. It specializes in analyzing transactions and assists in building specific models. Based on community detection algorithms， the system? analyzes general capital data and divides them into data clusters; by digging into the capital features of data clusters， we are able to sort out those relevant to suspicious criminal groups. With more detailed analysis of such clusters， we can narrow down our targets. Once we identify suspicious groups， we apply centrality algorithms and “dazzle” algorithms to further analyze the capital data of each accomplice for a more detailed picture. In this way， we can identify the main perpetrators and the organizational structure， developing insight into both the big picture of the case and each individual criminal.

The “Xingtu” system has successfully discovered 29 criminal groups from millions of financial data provided by the Economic Crime Investigation Bureau of the Public Security Ministry. During this process， we uncovered and reported a large three-level criminal group headed by a suspect surnamed Chen. This finding was commended by the national authorities and won us first place in the tax-related crime investigation contest at the national “Lunjian 2018” competition.

In conclusion， against the backdrop of rapid technological advancement， we put forward a new concept of building an economic investigation “brain” with AI technologies. In 2020， the WHECID expanded on recent practices to develop an AI Intelligent Identification Platform underpinned by knowledge graph technology and algorithms on machine and deep learning.

This platform consists of four components： the knowledge graph， the algorithm center， the model center， and the tool center. The knowledge graph is the key technology， while the others are built to support it. With a proper mix of algorithms， the platforms knowledge extractor draws from open domains such as common sense， specialized knowledge in the tax investigation field， and the experience of police officers， and store it in knowledge bases.

The platform can also perform independent learning. Here is how it works： the embedded knowledge-acquiring tool， based on algorithm models， will learn automatically the knowledge inductively and deductively accumulated in the data. Meanwhile， it can retrieve and analyze different types of data， such as capital， phone bills， business operations and logistics. By combining experience and data， the platform is capable of analyzing economic crime data from different perspectives.

We have named it an “intelligent identification” platform because it is intelligent in using experience and data. “Identification” implies that it can apply existing rules to data analysis and case investigations. The WHECID has conducted a comprehensive analysis of inflated invoice cases in the electronics industry on this platform， uncovering 11 companies committing such crimes. We identified criminal groups who altered invoices， acted as intermediaries， and provided fraudulent invoices upstream. We then carried out precision attacks， successfully concluding the “8·12” case in the mobile phone industry.

Looking ahead， do you have any plan for the future？

The WHECID will continue to make breakthroughs. As we apply more information technologies to handle cases， we are striving to create a “brain” for economic crime investigation capable of independent thinking， judgment， active discovery， and providing real-time warnings of criminal activity. By doing so， we can truly apply theories to practices， map out Wuhans evolving economic structure， closely follow its pillar industries， and guide and regulate enterprises development. We are endeavoring to become a barometer of Wuhans economic performance， improve our capacity to serve the enterprise development， and safeguard economic growth.

（Translated by Agnes）