中國軍隊在智慧戰場上進行作戰欺騙
現代英語：
It is easy to break the “fog” of the battlefield, but it is difficult to break the “obsession” in your heart——
Since ancient times, achieving surprise through combat deception has been an important way to win on the battlefield. Entering the era of intelligence, the in-depth application of artificial intelligence technology has not only clearly dispelled the original war “fog”, but also created a large amount of new war “fog”. If we only rely on improving deception techniques and means, and simply superimposing and strengthening the traditional deception paradigm, it will become increasingly difficult to achieve the deception goal. From “smart deception” to “smart victory”, there is an urgent need for an overall transformation of the objects of deception, means of deception, methods of deception, and focus of deception, so as to form a new deception paradigm that meets the requirements of the intelligent era.
The target of deception has shifted from humans to human-machine hybrid agents
Clausewitz believed that three-quarters of the factors on which war is based are more or less surrounded by the “fog” of uncertainty. Combat deception is essentially the use of uncertainty in war. The more “fog” there is in war, the more room there is for maneuvering. Traditional combat deception is carried out around the opponent’s decision-making level, and people are the only target of deception. However, with the increasingly prominent role of intelligent intelligence analysis and auxiliary decision-making systems in command activities, the use of deception to achieve strategic, campaign, and tactical surprises faces major challenges. How to deceive human-machine hybrid intelligent entities composed of humans and intelligent systems has become an important factor that needs to be considered when planning and implementing deception in the intelligent era. The competition surrounding intelligent deception and anti-deception is becoming increasingly fierce.
There is a world of difference between deceiving people and deceiving intelligent systems. In the past, the “calculations” that deceived people may be exposed when facing the “calculations” of intelligent systems. Intelligent systems can efficiently integrate and process massive amounts of sensor data and Internet open source information, making a qualitative leap in the speed, depth, breadth and accuracy of battlefield situation perception, realizing a profound transformation from “sensing” to “knowing”, from “state” to “momentum”, and playing an important role in dispelling the “fog” of war. For example, on the battlefield, although both sides try to hide the truth and cover up their intentions in various ways, they still cannot escape the “eyes” of the intelligent system: the tracks left by carefully disguised tanks and armored vehicles, after being detected by the opponent’s satellites, drones, etc., will also reveal their specific locations under the analysis of the intelligent system.
On the contrary, it is very easy to deceive intelligent systems with methods that target them, but it may not be able to deceive people. A foreign research team found that by changing a few key pixels in a picture of a cat, the intelligent system can identify the cat as a dog, while the human eye will not make any recognition errors due to this change. Similar incidents are common. Some studies have pointed out that sticking a piece of paper with a special pattern on a person’s forehead can deceive the strongest facial recognition system, and this method is highly portable and can deceive other facial recognition algorithms with a slight change.
It can be seen that deceiving people and deceiving intelligent systems are two different “deception methods”. After the deep application of artificial intelligence in the field of intelligence analysis and auxiliary decision-making, from the formulation of strategic deception plans to the design of battlefield camouflage patterns, how to deceive both the human brain and the computer and keep the human-machine hybrid intelligent body “in the dark” will be an important issue that needs to be focused on and solved in order to win the initiative in war.
The fraudulent methods have shifted from being mainly human-based to a combination of human and machine.
The organization and implementation of traditional combat deception is mainly manual, especially large-scale strategic deception, which requires a lot of manpower, material and financial resources. For example, in World War II, the Allies formulated a series of deception plans to ensure the success of the Normandy landing: setting up a fake radio network and a simulated landing fleet, and imagining that the US 1st Army Group with 50 divisions and 1 million people was actively preparing to cross the channel and land in the direction of Calais; using the air force to bomb Calais and Normandy, but the former was bombed more than 1 times more than the latter, etc. The application of artificial intelligence in deception can fundamentally change this situation. With humans as the main guide and intelligent means as the auxiliary, it can quickly generate massive amounts of false information, confusing the real with the fake, and create a thicker war “fog” for the opponent.
The use of intelligent means can improve the quality of deception. On the one hand, intelligent decision-making aids can be used to formulate deception plans, optimize the design of deception forces, deception deployment, deception processes, etc., to achieve systematic deception with the best overall effect; on the other hand, intelligent intelligence analysis systems can be used to pre-test the deception effect, “using one’s own spear to attack one’s own shield”, find out the loopholes and contradictions in the plan, and then improve the deception plan to make it logically self-consistent and seamless.
The use of intelligent means can expand the scale of deception. The increasingly mature deep fake technology can synthesize realistic fake pictures, handwriting, audio, video, etc. in large quantities, and has broad application prospects in strategic, campaign, and tactical deception. For example, in strategic campaign deception, corresponding technical means can be used to confuse opponents by forging fake radio stations and fake commanders, and even to fake an active command post in a certain battle direction; in tactical deception, battlefield camouflage can be used to attach special patterns to high-value equipment to make the opponent’s intelligent system recognize it incorrectly.
The use of intelligent means can reduce the cost of deception. With the support of technologies such as virtual reality and deep fakes, unexpected deception effects can often be achieved with the help of synthetic optics, acoustics and other means, and they are low-cost and low-investment, which is more cost-effective than traditional strategic deception methods. For example, setting up false targets such as bait unmanned combat platforms, using electronic feints and electronic camouflage to send false signals can effectively restrain the opponent’s power, produce high returns at low cost, and thus gain the upper hand.
The use of intelligent means can optimize the accuracy of deception. Traditional combat deception is usually stereotyped, with prominent characteristics of broadcast, extensive, and generalized. For this reason, in the era of intelligence, we should focus on collecting data on opponent decision makers in peacetime and use big data for precise analysis to “know the enemy” more deeply and specifically. On this basis, deep fake technology can be used in wartime to customize the content of deception, realizing precise deception from targeting groups to targeting individuals.
The method of deception has shifted from mainly deceiving to mainly confusing and seducing.
“Playing cards” and “playing chess” are two game modes with completely different battlefield transparency. In the “playing cards” mode, both sides only know the cards that the opponent has played, but do not know the cards in the opponent’s hand, let alone what cards the opponent will play next; while in the “playing chess” mode, the deployment of both sides’ forces on the chessboard is completely transparent, but the opponent’s intentions and the next move are unknown. It is not difficult to see that from cold weapon wars, hot weapon wars, mechanized wars, informationized wars, and then to intelligent wars, the form of war confrontation is increasingly changing from the “playing cards” mode to the “playing chess” mode.
In a war of “playing cards”, blind deception is very useful. Through strict disguise and strict confidentiality, the opponent’s channels of information can be blocked as much as possible, making it impossible for the opponent to detect one’s own intentions and actions, thereby achieving surprise. In the past, when the means of obtaining information were limited and information on the battlefield situation was scarce, there were many examples of wars that used “hiding the truth” and “showing falsehood” to achieve surprise. However, at present, with the help of advanced reconnaissance technology, full-dimensional and full-spectrum reconnaissance has been realized, and the battlefield is becoming more and more transparent. Complete concealment without any revealing features is difficult to achieve. Once the concealment state is switched to the action state, the probability of being discovered by the opponent will be greatly increased. Blind deception can only become an auxiliary deception method.
In the war of “chess”, the following two deception methods are usually used: one is confusing deception, that is, using intelligent means to send a large amount of true and false mixed and difficult to identify information, increasing the ambiguity of information and the difficulty of analysis, making it difficult for the opponent to judge or misjudge. The second is inducement deception, that is, by sending high-definition misleading information, the opponent is led into a preset trap. The combination of these two methods and the cooperation of blinding deception together constitute a hybrid deception that is difficult for the opponent to guard against.
The focus of deception shifts from human perception to human cognition
As the main subject of war, people are important variables that influence the war situation, which implies uncertainty and uncontrollability. From the perspective of psychology, cognitive neurology and other aspects, the “black box” of the mind still cannot be revealed. Deception by deception targets people’s eyes and ears, taking advantage of human sensory weaknesses, while deception by deception and temptation directly targets people’s minds, taking advantage of human weaknesses.
From past cases, even with the most advanced intelligence surveillance and reconnaissance technology and the most intelligent analysis methods, it is impossible to make up for and overcome human weaknesses. In many cases, it is not that the intelligence department failed to recognize the opponent’s deception, but that the decision-makers are unwilling to believe the facts. On the eve of the Soviet-German War in World War II, although more and more evidence showed that Germany was planning to invade the Soviet Union, the Soviet decision-makers believed that the war would not come for the time being. Therefore, when the war broke out, the Soviet army was not well prepared for the response, and the initial defensive actions were very passive.
War practice shows that in the era of intelligence, even if the opponent has obvious military technology advantages and can achieve one-way transparency on the battlefield through advanced intelligence surveillance and reconnaissance technology, the enemy can still take advantage of the cognitive weaknesses of the opponent’s decision-making layer to implement counter-intuitive deception and cover up the true intentions and actions. This also shows that the focus and center of deception in the era of intelligence should not be entirely on how to deliberately cover up the traces of military actions, but should focus more on targeting the opponent’s decision-making layer and inducing it to make decisions and actions that the enemy wants to see.
Yuan Yi Zhao Di
(Author’s unit: Institute of War Studies, Academy of Military Science)
現代國語：
破戰場“迷霧”易，破心中“執念”難——
袁 藝 趙 頔
自古以來，透過作戰欺騙達成突然性，是戰場制勝的重要途徑。進入智慧化時代，人工智慧技術的深度應用，在清晰撥開原有戰爭「迷霧」的同時，又製造出大量新的戰爭「迷霧」。如果只依賴改進欺騙技術和手段，在傳統欺騙範式上做簡單的疊加強化，就想達成欺騙目標的難度越來越大。由“智騙”到“智勝”，迫切需要欺騙對象、欺騙手段、欺騙方式、欺騙重心等各個方面的整體轉變，形成適應智能化時代要求的新型欺騙範式。
欺騙對象由人轉向人機混合智能體
克勞塞維茨認為，戰爭所依據的四分之三的因素或多或少被不確定性的「迷霧」包圍著。作戰欺騙本質上就是對戰爭中不確定性的利用，戰爭「迷霧」越多，施計用謀的空間就越大。傳統作戰欺騙都是圍繞著對方決策層而展開的，人是欺騙的唯一對象。但隨著智慧情報分析與輔助決策系統在指揮活動中的地位作用日益凸顯，以欺騙達成戰略、戰役、戰術突然性面臨重大挑戰。如何欺騙人與智慧系統共同組成的人機混合智能體，成為智能化時代籌劃實施欺騙需要考慮的重要因素，圍繞智能欺騙與反欺騙的較量日趨激烈。
欺騙人與欺騙智慧系統有著天壤之別，以往欺騙人的「算計」在面對智慧系統的「計算」時可能會被識破。智慧型系統可高效融合處理海量的傳感器數據和互聯網開源信息，使得戰場態勢感知的速度、深度、廣度和精度產生質的飛躍，實現由“感”到“知”、由“態”到“勢”的深刻轉變，在撥開戰爭「迷霧」方面發揮重要作用。例如，戰場上盡管交戰雙方都試圖用各種方法隱藏真相、掩蓋企圖，但仍逃不出智能係統的「慧眼」：精心偽裝的坦克、裝甲車等留下的車轍痕跡，被對方衛星、無人機等偵照後，在智慧型系統的分析下也會暴露出具體位置。
相反，針對智慧型系統的欺騙方式欺騙智慧系統非常容易，但可能又欺騙不了人。國外研究團隊發現，只要改變一隻貓的圖片中的少數幾個關鍵像素，就可以使智慧系統將貓識別為狗，而人眼則完全不會因這種變化而出現識別錯誤。類似的事件屢見不鮮，有研究指出，在人類前額上貼一張有特殊圖案的紙片，就能夠騙過最強的人臉識別系統，且這一方法具有很強的可移植性，稍加改變就可以欺騙其他的人臉識別演算法。
由此可見，欺騙人與欺騙智慧系統是兩種不同的「騙法」。人工智慧深度應用於情報分析與輔助決策領域後，大到戰略欺騙方案的製定，小到戰場迷彩圖案的設計，如何既騙過人腦又騙過電腦，把人機混合智能體「蒙在鼓裡”，將會是贏得戰爭主動權需要重點關注並加以解決的重要課題。
欺騙手段由人工為主轉向人機結合
傳統作戰欺騙的組織實施以人工為主，尤其是大規模的戰略欺騙，需要投入大量的人力物力財力。例如，二戰時盟軍為確保諾曼底登陸成功，制定了一系列疑兵計畫：建立假的無線電網和模擬登陸艦隊，虛構有50個師100萬人的美第1集團軍群，正在積極準備橫渡海峽向加萊方向登陸；使用空軍對加萊和諾曼底進行轟炸，但前者遭到的轟炸比後者多1倍以上等等。人工智慧運用於欺騙可從根本上改變這一局面，以人為主導輔以智能手段，可快速生成海量虛假信息，以假亂真，給對手製造更加濃厚的戰爭“迷霧”。
運用智慧手段可提升欺騙品質。一方面，可運用智慧輔助決策手段訂定欺騙計畫，優化設計欺騙力量、欺騙部署、欺騙流程等，實現全局效果最佳的體系化欺騙；另一方面，可運用智慧情報分析系統預先檢驗欺騙效果， “以己之矛攻己之盾”，找出計劃中的漏洞和矛盾點，進而完善欺騙計劃，使其邏輯自洽、嚴絲合縫。
運用智慧手段可擴大欺騙規模。日益成熟的深度偽造技術，可大量合成逼真的虛假圖片、筆跡、音頻、視頻等，在戰略、戰役、戰術欺騙中有著廣泛的應用前景。例如，在戰略戰役欺騙方面，可透過相應技術手段，偽造假電台、假指揮員等迷惑對手，甚至能夠在某一戰役方向偽造一個活躍的指揮所；在戰術欺騙方面，可通過戰場偽裝，給高價值裝備貼上特製圖案，使對手的智慧系統識別出錯。
運用智慧手段可降低欺騙成本。在虛擬現實、深度偽造等技術的支持下，借助合成光學、聲學等手段往往也能達到意想不到的欺騙效果，並且兼具低成本、小投入的特點，相比傳統戰略欺騙方式具有高效費比優勢。如設置誘餌無人作戰平台等假目標，運用電子佯動、電子偽裝等施放假信號，都能夠有效牽制對手力量，以低成本產出高回報，從而贏得制勝先機。
運用智慧手段可優化欺騙精度。傳統作戰欺騙通常千篇一律，廣播式、粗放式、概略化特點比較突出。為此，智能化時代，平時就應注重廣泛收集對手決策者數據，並運用大數據進行精確分析，以更加深刻更加具體地「知彼」。在此基礎上，戰時就可運用深度偽造技術個性化客製化欺騙內容，實現由針對群體到瞄準個體的精準欺騙。
欺騙方式由以蒙蔽為主轉向以迷惑、誘導為主
「打牌」和「下棋」是戰場透明度截然不同的兩種賽局模式。 「打牌」模式中，雙方都只知道對手已出的牌，但不知道對手手中的牌，更不知道下一步對手會出什麼牌；而「下棋」模式中，棋盤上雙方兵力部署完全透明，但不知道對手企圖和下一步棋怎麼走。不難看出，從冷兵器戰爭、熱兵器戰爭、機械化戰爭到資訊化戰爭，再到智慧化戰爭，戰爭對抗形式日益由「打牌」模式轉變為「下棋」模式。
在「打牌」模式的戰爭中，蒙蔽式欺騙非常管用，可通過嚴密偽裝和嚴格保密，盡可能地封鎖對手的獲情渠道，使其無法察覺己方企圖和行動，進而達成突然性。在過去資訊獲取手段有限、戰場態勢資訊匱乏的年代，主用「隱真」輔以「示假」達成突然性的戰例很多。但當前，憑借先進偵察技術，已經實現了全維全譜偵察，戰場透明化程度越來越高，無任何暴露特徵的完全隱蔽已難以實現，而一旦由隱蔽狀態轉入行動狀態，更會大大增加被對手發現的機率，蒙蔽式欺騙只能成為輔助欺騙手段。
在「下棋」模式的戰爭中，通常採用以下兩種欺騙方式：一是迷惑式欺騙，即藉助智能手段，發出大量真假混雜、難以辨認的信息，增大信息模糊度和分析難度，使對手難以判斷或判斷失誤。二是誘導式欺騙，即透過發出高清晰誤導訊息，將對手引入預設陷阱。兩種方式結合再加上蒙蔽式欺騙的配合，共同構成了對手難以防範的混合式欺騙。
欺騙重心由人的感知轉向人的認知
作為戰爭的主體，人是左右戰局的重要變量，蘊含著不確定性和不可控性。從心理學、認知神經學等層面來看，心智的「黑箱」仍然無法揭開。蒙蔽式欺騙針對的是人的耳目，利用的是人類感官弱點，而迷惑式和誘導式欺騙直指人的心智，利用的是人性弱點。
從以往案例來看，即使擁有最先進的情報監視偵察技術和最聰明化的分析手段，也無法彌補和克服人性弱點。很多情況下，不是情報部門沒有辨識出對手的欺騙，而是決策層不願意相信事實。在第二次世界大戰蘇德戰爭前夕，盡管當時越來越多的證據表明，德國正計劃入侵蘇聯，但蘇聯決策層認為戰爭暫時不會來臨，所以當戰爭爆發時，沒有做好應對準備的蘇軍，前期的防禦行動非常被動。
戰爭實踐表明，進入智能化時代，即使對手擁有明顯的軍事技術優勢，能夠通過先進的情報監視偵察技術達成戰場單向透明，但己方仍可利用對手決策層的認知弱點，實施反直覺欺騙，掩蓋真實意圖和行動。這也表明，智能化時代欺騙的發力點和重心，不應全部放在如何刻意掩蓋軍事行動痕跡上，而應更加註重針對對手決策層，誘導其作出己方希望看到的決策行動。
（作者單位：軍事科學院戰爭研究院）
